Email Policies

Disclaimers
Disclaimers
Pro Edition feature. Maps to Email Policies > Disclaimers (
view_disclaimers.cfm, 
edit_disclaimer.cfm, 
disclaimer_delete.cfm).
Hermes appends a configurable disclaimer to outbound mail at the gateway, with two scopes:
Scope
Sender match
Use case
Domain
All senders in 
@example.com
Default org-wide compliance/legal language
Relay Recipient
Specific full address (e.g. 
vendor@example.com)
Per-relay-user override for tenants with extra regulatory language
Most-specific match wins: a relay-recipient match is used before the domain default.
Pipeline placement
Disclaimers are applied at SMTP receive time by the 
hermes_body_milter container, which Postfix consults as a milter alongside OpenDKIM and OpenDMARC.
External MTA / MUA submission
        │
        ▼
Postfix smtpd
   ├─ smtpd_milters chain (in order):
   │   1. OpenDKIM            (signs/verifies)
   │   2. OpenDMARC           (DMARC policy)
   │   3. hermes_body_milter  (THIS — disclaimers, signatures, banners)
   ▼
content_filter → Amavis    (unmodified path; sees the body milter's output)
   ▼
Ciphermail              (server-side S/MIME or PGP, if configured)
   ▼
Postfix :10026          (OpenDKIM signs the final composed body here)
   ▼
external
Body modification happens at smtpd time, before content_filter routes to Amavis. By the time Amavis sees the message, the disclaimer is already baked in. Amavis processes a normal-looking message; no internal-state coupling, no temp-file races.
OpenDKIM's outbound signing fires at the 
:10026 re-injection — after both the body milter and Ciphermail. Hermes' own DKIM therefore always covers whatever the recipient ultimately receives. Ciphermail's server-side crypto also covers the disclaimer because Ciphermail runs after the milter.
Behavior with S/MIME, PGP, and DKIM-signed mail
The behavior depends on who signed/encrypted the message and when in the pipeline.
Server-side: signed/encrypted by Ciphermail — disclaimer is applied
Ciphermail runs after the body milter. Mail arrives at the milter as plaintext, the disclaimer is appended, then Ciphermail signs or encrypts the modified body. The recipient sees a valid signature and the disclaimer. No conflict.
Client-side: signed/encrypted by the user's MUA — disclaimer is skipped
Mail signed in Outlook (S/MIME) or Thunderbird+Enigmail (PGP) arrives at the gateway with the cryptographic envelope already sealed. Modifying the body would either invalidate the signature or mangle the ciphertext.
The body milter detects the following patterns in the headers (or first 32 KB of the body) and exits unchanged when any matches:
Pattern matched
Meaning
Content-Type: multipart/signed; protocol="application/pkcs7-signature"
S/MIME detached signature
Content-Type: application/pkcs7-mime
S/MIME opaque-signed or enveloped
Content-Type: multipart/signed; protocol="application/pgp-signature"
PGP/MIME detached signature
Content-Type: multipart/encrypted; protocol="application/pgp-encrypted"
PGP/MIME encrypted
-----BEGIN PGP SIGNED MESSAGE----- in body
PGP inline-signed
-----BEGIN PGP MESSAGE----- in body
PGP inline-encrypted
When any of those match, the body is left untouched, the signature stays valid, the user's legal-text expectations are preserved (their MUA template is already in the body), and the gateway gets out of the way.
Operational consequence. A site whose users sign client-side will not get gateway disclaimers on those specific signed messages — by design. If org-wide legal text on all outbound is mandatory, the only safe pattern is server-side signing in Ciphermail with the disclaimer applied first.
DKIM: Hermes-signed mail is fine; upstream-signed mail is skipped
OpenDKIM signs at the Postfix 
:10026 re-injection step — after the body milter. So Hermes' own DKIM signature always covers the recipient's view of the message (with disclaimer baked in). No conflict.
The risk is mail that arrives at Hermes already DKIM-signed by an upstream MTA — typically a relay user whose own mail server signs before forwarding through us. Modifying that body would invalidate the upstream signature at the recipient.
The body milter treats a pre-existing 
DKIM-Signature: header the same way as a sealed S/MIME or PGP envelope and skips the disclaimer. Since Hermes' own DKIM signs at 
:10026 (downstream of this milter), any DKIM-Signature header present at the milter's point in the pipeline came from somewhere upstream of Hermes.
Reply-chain handling — no dedup, by design
The milter does not detect or skip messages that already carry a previous disclaimer in their quoted history. Every outbound message gets a fresh disclaimer applied — including replies inside a long thread.
This matches industry norm: commercial server-side disclaimer / signature platforms (Exclaimer, Crossware, CodeTwo, Microsoft 365 transport rules) all stamp every outbound without dedup. The reasoning:
Compliance. Many regulatory regimes (HIPAA email confidentiality, GDPR data-controller notices, financial-services disclosure) treat each transmission as requiring its own disclaimer. Stamping only the first message in a thread arguably leaves later replies non-compliant.
Self-contained messages. If a recipient forwards a reply (with quoted history) to a third party, the disclaimer is preserved per-message in the forwarded text.
Predictable behavior. Operators don't have to explain "sometimes the disclaimer shows, sometimes it doesn't."
Cosmetic concern is weak. Modern MUAs (Gmail, Outlook, Apple Mail) collapse quoted history by default, so stacked disclaimers in long threads are rarely visible to readers.
Earlier iterations of #214 included a sentinel-marker dedup mechanism (
[HD] / 
<!-- HERMES_DISCLAIMER_V1 -->). That was removed during DEV testing in favor of the industry-norm pattern.
Position: append vs prepend
The schema and UI both expose 
position = append | prepend, but v1 honors append only. Prepend is tracked as a v2 enhancement.
Failure semantics
The body milter is graceful-degradation by design. Postfix's 
milter_default_action = accept means:
Milter container down or unreachable → mail flows unmodified (missed disclaimer, but no delivery outage)
Map file unreadable → no entries match → all mail flows unmodified
Modifier raises an exception → caught and logged → mail flows unmodified
altermime / parse errors → caught and logged → mail flows unmodified
In every failure case, mail keeps flowing. Worst case is a missed disclaimer, never lost mail. Compare the legacy "modify in amavis hook" approach (#214 Phase 3 v1, retired) which silently dropped messages when the in-place body modification desynced amavis's internal state.
Files generated on save/delete
The CFML include 
inc/disclaimer_write_and_reload.cfm runs after every save or delete and rewrites the entire on-disk state from the 
disclaimers table:
/etc/hermes/body_milter/disclaimers/disclaimer_by_sender   sender → option map
/etc/hermes/body_milter/disclaimers/files/<option>/
    body.txt          plain-text disclaimer
    body.html         html disclaimer (may have <img src="cid:..."> refs)
    images/
        1.png         per-disclaimer inline images (#230)
        2.jpg
        ...
Where 
<option> is 
domain_<safe> or 
relay_<safe> (non-alphanumeric chars in the source key are replaced with 
_).
Each disclaimer gets its own subdirectory. The files directory is wiped (per-option subdirectories deleted recursively, but the parent 
files/ directory and its 
.gitkeep are preserved) and rewritten on every save. There is no incremental update — this guarantees deleted rows and renamed scope keys never leave stale files (or stale image binaries) behind.
No reload step needed. The body milter mtime-watches each map file on every message and reloads when it changes. The CFML 
cffile write to the map file is enough to make the change take effect on the next message processed by the milter.
Inline images (#230)
Admins can paste or upload images directly into the Quill editor when authoring a disclaimer. Supported formats: PNG, JPEG, GIF. SVG and WebP are explicitly rejected (security and recipient-compatibility reasons). Limits enforced at save time:
5 images max per disclaimer
200 KB per image (after base64 decode)
1 MB total across all images in a single disclaimer
If any limit is exceeded, the save is rejected with a specific error explaining what failed. Admins can reduce image count or size and re-save.
How it works:
Quill embeds pasted/uploaded images as base64 inline 
<img src="data:image/...;base64,..."> in the HTML body. The base64 representation is what's stored in the 
disclaimers.body_html column.
At save time, the regenerator parses 
body_html for 
data: URLs, decodes each base64 blob, writes the binary as 
<option>/images/<N>.<ext>, and rewrites the HTML in 
<option>/body.html to use 
<img src="cid:disclaimer_<option>_img_<N>"> references.
At message-send time, the body milter reads 
body.html, walks 
<img src="cid:..."> references, and attaches each referenced image as an 
image/<format> MIME part with 
Content-ID: <disclaimer_<option>_img_<N>> and 
Content-Disposition: inline.
The milter wraps the message as 
multipart/related so the recipient MUA resolves cid references against the inline parts.
MIME structure transformation (representative example):
Original outbound:
  multipart/alternative
    text/plain
    text/html (no images)
After milter (with disclaimer including 1 image):
  multipart/related
    multipart/alternative
      text/plain  (with text disclaimer appended; images omitted from text)
      text/html   (with html disclaimer + <img src="cid:...">)
    image/png
      Content-ID: <disclaimer_..._img_1>
      Content-Disposition: inline
This structure renders inline in all major MUAs (Gmail, Outlook, Apple Mail, Thunderbird, mobile clients).
The plain-text version of the disclaimer omits images entirely — base64 inline images don't translate to text, and recipients viewing the message in plain-text mode see the disclaimer text without any image markers.
Hermes' own DKIM signature covers the modified body (including the multipart/related wrap and image parts), because OpenDKIM signs at the postfix 
:10026 re-injection step — downstream of the body milter. The signature validates against what the recipient receives.
Auto-derive of plain-text part
The Quill editor on 
edit_disclaimer.cfm drives 
body_html. By default the plain-text part shipped to recipients with a non-HTML MUA is auto-derived from the HTML on save: 
<br>, 
</p>, 
</li> become newlines, all other tags are stripped, runs of 3+ newlines collapse to 2.
Admins who need character-perfect plain text different from the auto-strip (e.g. for regulated industries) can toggle Edit plain-text version separately to expose a second editor. When set, 
body_text is shipped verbatim instead of derived.
Disabled rows
Rows with 
enabled = 0 are skipped entirely on regen — no files written, no map entry. The milter never matches that scope until the row is re-enabled.
Internal-only mail
v1 does not suppress disclaimers for internal-only mail (sender + all recipients in 
@local_domains). Domain disclaimers will be applied to internal mail in the same domain. If this is a problem for your install, file a feature request to add an internal-only bypass.
Why a separate milter and not an amavis hook
Earlier #214 iterations attempted to dispatch the disclaimer from inside an amavisd-new 
Custom.pm 
before_send hook, calling altermime via 
system() on the temp file amavis was managing. amavisd-new 2.13 caused two problems: the legacy 
@disclaimer_options_bysender_maps dispatch path was removed (variables still parse but no code reads them), and the 
before_send hook documentation says "may modify mail" but in practice in-place body modification desynchronizes amavis's internal MIME state and silently loses mail.
The body milter approach moves the body-modification step out of amavis entirely. amavis's role is unchanged from before #214 ever existed; the milter sits in postfix's smtpd_milters chain alongside OpenDKIM and OpenDMARC, the same architectural pattern Hermes already uses for body-touching policy enforcement. amavis is fully decoupled from the disclaimer feature, which means amavis upgrades and the disclaimer feature evolve independently.
This same milter container is intended to host:
#226 User Signatures (per-mailbox personal text from LDAP attributes or user-portal editor)
#228 External Sender Banner (warning banner on inbound external mail)
Future Link Guard (URL rewriting through a click-through endpoint)
Each is a 
Modifier subclass in 
/usr/local/bin/hermes-body-milter registered in the 
MODIFIERS list. The dispatcher is unchanged.

External Banner
External Banner
Maps to Email Policies > External Banner (
view_external_banners.cfm, 
edit_external_banner.cfm, 
external_banner_delete.cfm). Available on both Community and Pro editions — phishing protection is a baseline security feature, not a Pro upsell.
Hermes prepends (or optionally appends) a warning banner to inbound mail from external senders destined for a local recipient. The banner is injected into the message body itself, so every MUA — webmail, Outlook, Apple Mail, mobile clients — renders it without relying on transport rules or recipient-side configuration. Tracked as #228.
Scope
Scope
Recipient match
Use case
System default
All recipient domains (no override)
Single banner used everywhere; recommended starting point
Per-recipient-domain
Specific local mailbox domain (e.g. 
legal.example.com)
Different copy or compliance language for one domain
Resolution at message time, in the body milter's 
ExternalBannerModifier:
Look up the first local recipient's domain in 
/etc/hermes/body_milter/banners/banner_by_recipient_domain.
If a matching row exists, use it.
Otherwise fall back to the 
_default system-wide entry.
Otherwise no banner is applied.
Only the first local recipient is consulted — mixed-domain envelopes get the banner of the first local recipient encountered. This keeps the modification deterministic regardless of envelope ordering.
The 
recipient_domain field is locked after creation. Delete and re-create the row to change scope.
What counts as "external"
The body milter uses Postfix's 
/etc/postfix/relay_domains file as the source of truth for "local". A message is considered inbound from an external sender when:
The 
MAIL FROM sender domain is not in 
relay_domains, AND
At least one 
RCPT TO recipient domain is in 
relay_domains.
Internal-to-internal mail (sender + all recipients local) is classified as 
direction = internal and the banner is not applied. There is no separate allowlist of "trusted partner" external senders today — every external sender to a local recipient triggers the banner if one is configured for that recipient's domain.
Pipeline placement
The banner is injected at SMTP receive time by the 
hermes_body_milter container, the same container that emits outbound disclaimers (disclaimers.md) and organizational signatures (organizational-signatures.md). The milter listens on 
inet:hermes_body_milter:8893 and Postfix consults it as part of 
smtpd_milters.
Inbound external MTA
        |
        v
Postfix smtpd
   +- smtpd_milters chain (in order):
   |    1. OpenDKIM            (verifies upstream DKIM signature)
   |    2. OpenDMARC           (DMARC policy + ARC verification)
   |    3. hermes_body_milter  (THIS -- banner prepended here)
   |       --> Authentication-Results header has already been written
   |           by OpenDKIM/OpenDMARC BEFORE the banner touches the body
   v
content_filter --> Amavis    (sees the banner-prepended body)
   v
Ciphermail              (server-side S/MIME or PGP, if configured)
   v
Postfix :10026          (multi-instance OpenDKIM re-signs the final body)
   v
Local delivery (Dovecot LMTP)
Key ordering points:
OpenDKIM verifies first. The upstream sender's DKIM verdict is captured in 
Authentication-Results: headers before the banner is injected. The header is preserved on the message; the banner does not retroactively change what OpenDKIM saw at smtpd time.
Amavis sees the modified body. Spam scoring runs against the banner-prepended message. This is intentional — the banner content is short and stable and does not skew SpamAssassin scores in practice.
Hermes' downstream re-sign covers the modified body. The multi-instance OpenDKIM at 
:10026 (#232) signs after Ciphermail rebuild, so the final outgoing-to-Dovecot body is covered by Hermes' own signature.
Behavior with signed and encrypted mail
The modifier inherits the same skip rules as Disclaimers for sealed envelopes:
Pattern matched
Meaning
Banner action
Content-Type: multipart/signed; protocol="application/pkcs7-signature"
S/MIME detached
Skip
Content-Type: application/pkcs7-mime
S/MIME opaque/enveloped
Skip
Content-Type: multipart/signed; protocol="application/pgp-signature"
PGP/MIME detached
Skip
Content-Type: multipart/encrypted; protocol="application/pgp-encrypted"
PGP/MIME encrypted
Skip
-----BEGIN PGP SIGNED MESSAGE----- in body
PGP inline-signed
Skip
-----BEGIN PGP MESSAGE----- in body
PGP inline-encrypted
Skip
Pre-existing 
DKIM-Signature: header on inbound mail
Upstream DKIM signed
Modify anyway (see below)
The corresponding flags on 
ExternalBannerModifier are 
skip_on_signed = True, 
skip_on_pgp_inline = True, 
skip_on_dkim = False.
Why the banner does NOT skip on upstream DKIM
About 95% of inbound mail today carries a 
DKIM-Signature: header. If the banner skipped on DKIM, the feature would be effectively inert — the warning would only land on the unsigned minority that needs it least.
Hermes already records the upstream DKIM verdict in 
Authentication-Results: before modifying the body. Recipients overwhelmingly read mail through Dovecot/IMAP and the recipient MUA does not re-verify upstream DKIM. The banner is therefore safe in the common case.
The narrower edge case — a recipient who forwards Hermes-banner'd mail to a downstream MX that does re-verify upstream DKIM — is addressed by ARC sealing (#229). Hermes' ARC seal at 
:10026 records 
cv=fail for the upstream chain (because we modified the body), but the seal itself is mathematically valid and the downstream MX can trust Hermes' ARC verdict if Hermes is on its allowlist. See ARC Settings for the full discussion of the 
cv=fail-by-design pattern.
Operational consequence. Banner injection breaks the original sender's DKIM body hash and any upstream ARC body hash. This is by design. Hermes is the authoritative auth boundary for the domains it relays; customer downstream MX servers must allowlist Hermes and accept its delivered mail without re-running DKIM/SPF/DMARC/ARC. A downstream MX that re-verifies upstream auth on mail Hermes forwards is misconfigured — cross-ref ARC Settings, DKIM Settings, and DMARC Settings.
Position: prepend vs append
Position
Behavior
Recommendation
Top (
prepend)
Banner becomes the first child of the message body (above any quoted history)
Industry standard — users see the warning before reading any content
Bottom (
append)
Banner is appended after the user-visible body
Available for sites that prefer it; rarely used
Both positions are implemented end to end (unlike Disclaimers, where only 
append is honored in v1). HTML prepend is done with BeautifulSoup: the banner fragment is inserted as the first child of 
<body> when present, otherwise prepended to the root.
Templates
Banners use a server-side template gallery, not a free-form WYSIWYG editor. Quill 2.x's HTML normalization strips inline styles that Gmail and Outlook need (the same problem hit on Organizational Signatures #226 Phase 2 and on this feature), so admins pick a template and fill in form fields; the server renders pixel-perfect HTML at save time.
Bundled templates (each 
inc/external_banner_templates/<key>.cfm):
Template key
Display name
When to pick it
warning_yellow
Warning Yellow
Default. Yellow background with orange accent. Matches Microsoft 365 / Mimecast banner style most users recognize
critical_red
Critical Red
Red background, white text. Phishing-prone industries or post-incident periods where alert level needs to be raised
subtle_info
Subtle Info
Light gray with blue accent. Less alarming for high-volume inbound (support/sales) where alert fatigue is a concern
plain_text
Plain Text
Bold prefix + text, no background or border. Maximum cross-MUA compatibility, including text-only clients
All four templates expose the same field set:
Field
Type
Default
Notes
prefix
text
[EXTERNAL]
Short tag rendered bold at the start. Plain ASCII recommended for Outlook
headline
text
"This message originated from outside your organization."
First line, regular weight
body
text
"Do not click links or open attachments unless you recognize the sender..."
Second line, smaller text
show_learn_more
checkbox
false
Reveals the next two fields
learn_more_url
url
empty
Optional link to internal phishing-awareness training or wiki
learn_more_label
text
"Learn more about phishing"
Visible label for the learn-more link
All templates emit table-based HTML with 
bgcolor= attributes so Outlook (which strips inline CSS but honors deprecated HTML attributes) renders the banner correctly. Inline styles are belt-and-suspenders for Gmail, Apple Mail, and mobile clients.
The edit page renders a live preview in an iframe via 
inc/render_external_banner_preview.cfm so the admin sees exactly what 
save_external_banner_action.cfm will store.
Files generated on save/delete
inc/external_banner_write_and_reload.cfm runs after every save or delete and rewrites the entire on-disk state from the 
external_banners table:
/etc/hermes/body_milter/banners/banner_by_recipient_domain
    <recipient_domain>\t<option>
    _default\t<option>          special key, system-wide fallback
/etc/hermes/body_milter/banners/files/<option>/
    body.txt          plain-text banner (auto-derived at save)
    body.html         pre-rendered html banner
    position          "prepend" or "append" sidecar file
    images/           per-banner inline images (#230 cid pattern)
        1.png
        2.jpg
        ...
Where 
<option> is:
banner_default for the system-wide row (NULL 
recipient_domain)
banner_<sanitized_recipient_domain> for per-domain overrides (non-alphanumeric characters replaced with 
_)
The 
files/ subdirectory is wiped on every regen (per-banner subdirs deleted recursively; the 
.gitkeep is preserved). This guarantees deleted rows and renamed scopes never leave stale files behind.
No reload step needed. The body milter mtime-stats each map file on every message and reloads automatically when its mtime changes. The CFML 
cffile write to the map file is enough to make the change take effect on the next message.
Plain-text part
The HTML body stored in 
external_banners.body_html is rendered server-side from the chosen template. The plain-text counterpart in 
body_text is auto-derived at save time:
<br> becomes a newline
</p>, 
</li>, 
</tr>, 
</td>, 
</div> become newlines
All remaining tags are stripped
Runs of 3+ newlines collapse to 2
The plain-text version is shipped to recipients viewing the message as 
text/plain. Inline images are omitted from the plain-text part — data URLs don't translate to text and recipients in text mode see the banner copy without image markers.
Inline images (#230)
The banner modifier inherits the 
#230 cid inline-image pattern from Disclaimers. If a template's HTML contains 
<img src="cid:banner_<option>_img_<N>"> references, the body milter:
Loads matching 
images/<N>.<ext> files from the option directory.
Attaches each as an 
image/<format> MIME part with 
Content-ID: <banner_..._img_N> and 
Content-Disposition: inline.
Wraps the message as 
multipart/related so MUAs resolve cid references against the inline parts.
The cid prefix is 
banner_ so banner images cannot collide with 
disclaimer_ or 
signature_ cids inside the same composed message (the three modifiers can all add images to the same outbound; namespacing keeps them separate).
The bundled templates do not currently use inline images — banners are pure text. The infrastructure is present for future template additions (logo, warning icon, etc.).
Failure semantics
The body milter is graceful-degradation by design. Postfix's 
milter_default_action = accept means:
Milter container down or unreachable -> mail flows unmodified (missed banner, no delivery outage)
Map file unreadable -> no entries match -> all mail flows unmodified
Per-option files missing -> log + skip the modify -> mail flows unmodified
MIME parse exception -> caught and logged -> mail flows unmodified
Modifier raises any other exception -> caught and logged -> mail flows unmodified
In every failure case, mail keeps flowing. Worst case is a missed banner, never lost mail. Compare the legacy "modify in amavis hook" approach (#214 Phase 3 v1, retired) which silently dropped messages when the in-place body modification desynced amavis's internal state.
Disabled rows
Rows with 
enabled = 0 are skipped entirely during regen — no files written, no map entry. The milter never matches that scope until the row is re-enabled. Useful for staging copy changes before going live (build the new row disabled, preview it on 
edit_external_banner.cfm, flip the switch when ready).
Schema
CREATE TABLE IF NOT EXISTS external_banners (
  id               int(11)         NOT NULL AUTO_INCREMENT,
  recipient_domain varchar(255)    DEFAULT NULL,                -- NULL = system default
  template_key     varchar(64)     NOT NULL DEFAULT 'warning_yellow',
  fields_json      longtext        DEFAULT NULL,                 -- form values for re-edit
  body_text        longtext        DEFAULT NULL,                 -- auto-derived plain text
  body_html        longtext        NOT NULL,                     -- pre-rendered html
  position         enum('prepend','append') NOT NULL DEFAULT 'prepend',
  enabled          tinyint(3)      NOT NULL DEFAULT 1,
  updated_at       timestamp       NULL DEFAULT current_timestamp() ON UPDATE current_timestamp(),
  PRIMARY KEY (id),
  UNIQUE KEY uk_recipient_domain (recipient_domain)
);
The 
UNIQUE KEY on 
recipient_domain ensures only one row per recipient domain (and at most one system-default row where 
recipient_domain IS NULL). The 
fields_json blob stores the original form values so reopening the editor restores exactly what the admin typed; 
body_html is the rendered output the milter actually ships.
Verifying it works
The banner appears in the message body, so the easiest verification is to send an inbound message from an external account to a local mailbox and view the result in any MUA (webmail, Outlook, Apple Mail). Beyond that:
Body milter logs — the modifier logs 
external_banner applied: option=<name> position=<prepend|append> plain=<n> html=<n> per modified message. Surface with 
docker logs hermes_body_milter or via System Logs.
Authentication-Results: header is preserved from upstream and visible in the recipient's "view source"; this confirms OpenDKIM ran before the banner.
ARC-Seal: ... cv=fail in the outgoing message confirms the body was modified after the upstream chain — expected behavior, cross-ref ARC Settings.
Related
Disclaimers — the outbound counterpart; same 
hermes_body_milter container, parallel design (sender-keyed instead of recipient-keyed)
Organizational Signatures — second outbound modifier in the same container, with per-recipient resolution
ARC Settings — full explanation of 
cv=fail after body modification and the Hermes-as-auth-boundary model
DKIM Settings, DMARC Settings — upstream-verdict context preserved in 
Authentication-Results
Domains — local mailbox-hosting domains drive the per-domain dropdown on 
edit_external_banner.cfm
System Logs — surface the body-milter log stream for troubleshooting

Link Guard
Link Guard
Pro Edition feature. Maps to Email Policies > Link Guard (
view_linkguard.cfm, 
inc/linkguard_write_and_reload.cfm).
Link Guard provides time-of-click protection for inbound mail. Links in delivered messages are rewritten to point at a Hermes redirect endpoint; when a recipient clicks, Hermes evaluates the destination's reputation at that moment and decides — instantly — whether to allow, warn, or block. This closes the gap that delivery-time scanning misses: a link that is clean when the message arrives but is weaponized hours or days later.
Protection travels with the link. It works in the inbox, after a forward to a colleague, or when the message is opened days later on a phone — because the verdict is computed on click, not on delivery.
Components
Link Guard spans three containers plus the admin console:
Component
Role
hermes_body_milter
Rewrites inbound links at SMTP receive time (
LinkGuardModifier) and restores original links on outbound replies/forwards (
LinkGuardRestoreModifier).
hermes_linkguard
The verdict + redirect engine. Serves the public 
/lg/ click endpoint and a console-only management API. Holds the operational SQLite store (verdict cache, feeds, click log).
hermes_nginx
Reverse-proxies 
/lg/ from the public console host to the Link Guard container's public port.
Admin console
view_linkguard.cfm page; on save, 
inc/linkguard_write_and_reload.cfm pushes settings, scope, URL rules, and HMAC keys to the engine and reloads the milter maps.
The Link Guard container exposes a hardened two-port split:
Port
Surface
Exposure
8894 (public)
GET /lg/?t=<token>, 
POST /lg/proceed, 
GET /healthz
nginx-proxied; reachable by recipients clicking links
8895 (mgmt)
POST /api/config, 
POST /api/keys, 
POST /api/feed-refresh, 
GET /api/stats
console-only; never exposed publicly
The public surface can never push config or read keys; the management surface is never reachable from the internet.
Pipeline placement
Link Guard's two body-modification steps run in the 
hermes_body_milter container, the same milter Postfix consults for disclaimers, signatures, and banners.
INBOUND (rewrite)                          OUTBOUND (restore)
External MTA ──► Postfix smtpd             User reply/forward ──► Postfix smtpd
        │                                          │
        ▼                                          ▼
   smtpd_milters chain:                       smtpd_milters chain:
     1. OpenDKIM                                1. OpenDKIM
     2. OpenDMARC                               2. OpenDMARC
     3. hermes_body_milter                      3. hermes_body_milter
          └─ LinkGuardModifier:                      └─ LinkGuardRestoreModifier:
             rewrite links ──► /lg/ token              unwrap /lg/ tokens ──► original URLs
        │                                          │
        ▼                                          ▼
   Amavis ──► Ciphermail ──► deliver           Amavis ──► deliver to external
Rewriting happens at smtpd time, before content filtering. Hermes' own DKIM signs at the Postfix 
:10026 re-injection (downstream of the milter), so the signature always covers the rewritten body the recipient receives. Inbound mail that arrives already DKIM-signed, S/MIME-signed, or PGP-sealed is skipped — the same envelope-detection logic the disclaimer feature uses — so Link Guard never breaks an existing signature.
The click flow
Recipient clicks rewritten link
        │
        ▼
GET https://<console-host>/lg/?t=<token>   (nginx ──► linkguard :8894)
        │
        ├─ token invalid / expired ──► block page
        ▼
   resolve original URL from token
        │
        ▼
   verdict pipeline (see below) ──► {clean | suspicious | malicious}
        │
        ▼
   admin action for that tier:
     clean      ──► 302 redirect to the real URL   (default)
     suspicious ──► warning interstitial (resolved host shown; user may proceed)
     malicious  ──► block page (hard block, or block_override allowing proceed)
Every click is logged (recipient domain, URL hash, resolved host, verdict, source, action taken, client IP) for the reporting dashboard.
Verdict pipeline
verdict.resolve(url, recipient_domain) evaluates layers in precedence order and returns the first match:
#
Layer
Result
Notes
1
Admin blocklist
malicious
Operator-curated, console-managed
2
Admin allowlist
clean
Operator-curated; trumps feeds and heuristics
3
Verdict cache
cached result
Avoids re-running heuristics / re-hitting external APIs
4
Open-redirect extraction
escalate
Embedded target in an open-redirect param / nested URL is resolved and inherits its verdict (no fetch)
5
Local feeds
malicious
URLhaus / OpenPhish, stored in SQLite; only ever escalate
6
Heuristics
suspicious
Lookalike/punycode, IP-literal host, 
@ in authority, known shorteners, excessive subdomains, abused cloud-storage/redirector hosts
7
GSB / VirusTotal
malicious
Optional, admin-supplied keys; string-reputation lookups, cached. Also escalates a warn-tier (step 6) link to a hard block when flagged
8
Guarded redirect-follow
escalate
Optional (admin toggle): follow the 
30x chain under SSRF guards, verdict the final destination; also runs proactively on a step-6 warn so a malicious destination becomes a hard block
9
Default
clean
Nothing flagged it
Steps 1–7 never fetch the target URL — every reputation check sends the URL only as a string (Google Safe Browsing, VirusTotal). The only layer that makes an outbound request is the optional guarded redirect-follow (step 8), and only when an admin enables it; every hop is SSRF-fenced (see Redirect detection below). Local feeds only ever escalate a link to malicious; they never auto-allow.
URL shorteners are flagged suspicious (warn), not blocked — a shortener hides its real destination, which is exactly what time-of-click protection exists to surface. The warning interstitial shows the resolved host so the user can make an informed choice. The shortener set is a curated list of generic, anyone-shortens-anything services (referenced from the maintained 
PeterDaveHello/url-shorteners list); branded corporate shorteners (
aka.ms, 
amzn.to, 
a.co, 
adobe.ly, …) are deliberately excluded so legitimate branded short links don't warn-fatigue users. When 
follow_redirects is enabled, the real destination of any shortener is resolved regardless. VirusTotal requires ≥2 vendors flagging a URL before it counts as malicious, to cut false positives.
Verdict tiers and actions
Each verdict tier maps to an admin-configurable action:
Tier
Setting
Default action
Behavior
clean
action_clean
redirect
302 straight to the destination
suspicious
action_suspicious
warn
Interstitial; user may proceed
malicious
action_malicious
block
Block page
Available actions: 
redirect / 
allow (pass through), 
warn (interstitial with proceed), 
block (hard block, no proceed), 
block_override (block page that allows an explicit proceed).
A hard 
block can never be bypassed. 
POST /lg/proceed re-resolves and re-authorizes the verdict server-side — only a 
warn tier, or a 
block_override tier with the override flag, is allowed to continue. A user cannot escape a hard block by replaying the proceed request.
Redirect detection
Attackers increasingly chain through reputable hosts — 
storage.googleapis.com, 
firebasestorage.googleapis.com, 
*.web.app, Azure 
*.blob.core.windows.net, Cloudflare 
*.r2.dev / 
*.pages.dev, and classic open redirectors like 
google.com/url?q=… — because the host reputation is clean, so a string-only check passes the link through. Link Guard adds three layers to catch this "living off trusted services" pattern.
Open-redirect extraction (always on, no fetch)
The engine scans a link's query and fragment (raw, once-, and twice-URL-decoded) for an embedded 
http(s):// target on a different host than the redirector — the real destination of 
…/url?q=https://evil.example. That embedded target is re-run through the verdict pipeline (string-only); if it is suspicious or malicious, the original link inherits the verdict (
source = redirect, detail 
open-redirect -> <host>). A benign embedded URL is a no-op. No outbound request is made.
Abused-host heuristic (always on, no fetch)
Cloud-storage and app-hosting hosts commonly abused to host or bounce to phishing are flagged suspicious (warn) in the heuristics layer, so the user gets the interstitial and the resolved host instead of a silent redirect. Gated by 
flag_cloud_storage (default on). The match is suffix-based (
host == suffix or 
host ends with 
.suffix).
The list is an operator-managed table (
linkguard_abused_hosts), shipped pre-seeded with a curated baseline — there is no authoritative machine-readable feed of "abused hosting platforms" (the actual-bad URLs are what the URLhaus / OpenPhish feeds cover), so the seed is hand-curated from public abuse research (Trustwave SpiderLabs, Proofpoint, Netskope, Phishing.Database). It covers object storage (
storage.googleapis.com, 
firebasestorage.googleapis.com, 
firebaseapp.com, 
web.app, 
appspot.com, 
blob.core.windows.net, 
s3.amazonaws.com), edge/static-site hosting (
r2.dev, 
pages.dev, 
workers.dev, 
github.io, 
netlify.app, 
vercel.app, 
herokuapp.com, 
onrender.com, 
surge.sh), free site builders (
weebly.com, 
wixsite.com, 
000webhostapp.com), and tunneling services (
trycloudflare.com, 
ngrok.io, 
ngrok-free.app).
Manage it from the Abused / redirector hosts card on the Link Guard page:
Add any host (covers its subdomains) — react to a new abuse pattern immediately, no rebuild.
Remove a baseline entry you don't want (select → Delete). The seed is one-time (only when the table is empty), so your add/deletes survive upgrades.
Or suppress without deleting by adding the host/path to the admin URL allowlist — the allowlist wins over the heuristic (e.g. a company that legitimately serves files from 
storage.googleapis.com/<your-bucket>).
Add/delete in this card applies immediately — the handler regenerates the engine's 
config.json on each change (same as the URL-rules and domains cards), so there's no separate Save for table edits. The 
flag_cloud_storage master switch and the other settings live in the settings card and take effect on its Save & reload settings button.
Guarded redirect-follow (optional, the one fetch layer)
When 
follow_redirects is enabled, the engine — at click time, after the string layers — follows the HTTP 
30x redirect chain and verdicts the real destination, catching a trusted-host link that issues a server-side redirect to phishing (the 
storage.googleapis.com → phishing case). This is the only layer that makes an outbound request, so every hop is fenced (
_safe_to_fetch):
http/https only, and only a standard web port (80/443).
Each hop's host must resolve to public IPs only — any answer in a loopback / RFC1918 / link-local / reserved / multicast range aborts the follow. This stops the follower being used as an SSRF pivot into the internal Docker network.
HEAD only, no body downloaded; bounded by 
follow_max_hops (default 5), a short per-hop timeout, and a loop guard.
No cookies/credentials sent; neutral User-Agent; 
Referrer-Policy: no-referrer.
Each followed hop is verdicted string-only (a follow never recurses into another follow). If the chain reaches a suspicious/malicious destination, the link inherits that verdict (
source = redirect, detail 
redirect chain -> <final host>). A follow failure (timeout, guard stop, HEAD not allowed) fails closed for the follow — the link keeps whatever verdict the string layers produced; a click is never blocked by a follow error.
Proactive on warn-tier links. A shortener or abused-host link is flagged suspicious by the heuristics layer (step 6), which is before the follow layer in precedence. When 
follow_redirects is on, those warn-tier links are also followed at that point — proactively, during the verdict, not on "Continue":
destination resolves malicious → the link is escalated to a hard block (the user never gets a Continue into a known-bad site);
destination resolves clean or suspicious, or the chain can't be followed → the link stays a warn, now annotated with the resolved final host (
<reason> (resolves to <final host>)).
So with redirect-following enabled, a 
storage.googleapis.com / shortener link that actually bounces to phishing becomes a block, while one that resolves somewhere benign stays an informative warn. The verdict (block or warn) is cached, so the follow runs at most once per link per cache-TTL.
SSRF posture. Steps 1–7 never fetch the target. Enabling 
follow_redirects is a deliberate trade: it resolves redirect chains a string check cannot, at the cost of one guarded outbound request per uncached click (latency) and a controlled egress surface. The residual DNS-rebinding window (resolve-then-connect) is accepted for this release. Leave it off to preserve the zero-fetch guarantee; the two no-fetch layers above still run.
Out of scope for this release: following JavaScript or 
<meta> refresh redirects, which require fetching and parsing the page body — tracked as a later enhancement.
Tokens — stateful v2 (default) with stateless v1 fallback
The rewritten link carries a token in the 
t query parameter. Two formats exist:
v2 — stateful (default). The token is just 
2.<128-bit opaque id>. The milter writes the mapping 
id → {original_url, recipient_domain, expiry} to a shared SQLite store (
url_map.db) on the 
linkguard_data volume; the Link Guard container reads it. Because the token itself is tiny, there is no link-length limit — every link is protected regardless of how long the original URL is. This closes the v1 over-length fail-open gap (see below).
v1 — stateless (fallback + in-flight). The token is a self-contained HMAC signature: 
1.<recipient_domain>.<url>.<expiry>.<signature>. The milter mints a v1 token if the shared store is unavailable (e.g. off-box deployment, or transient DB contention), so mail flow never depends on the store. v1 tokens already in delivered mailboxes continue to verify until they age out via the token TTL.
The milter's mint/verify logic is a byte-for-byte mirror of the container's 
lg_token.py, so the container verifies exactly what the milter mints. The 
url_map.db store uses a rollback journal (not WAL), so the container can read it cross-container without a 
-shm file.
Why v2 exists. Under v1, a URL longer than the inline cap was left unprotected (the original link was passed through unrewritten). An attacker could pad a URL past the cap to dodge Link Guard entirely. v2's short opaque id removes the length dependency, so nothing is ever skipped. The 
max_inline_url setting is now a fallback-only bound for the v1 path.
Outbound link restoration
restore_outbound (default on) unwraps Link Guard tokens back to the original URLs on outbound mail — when a recipient replies to or forwards a protected message, the quoted history shows the real links again, not 
/lg/?t=... redirects. This keeps conversations readable and prevents Hermes redirect URLs from leaking to external parties. (Microsoft 365 was verified not to strip the tokens on manual replies, so restoration is the correct default.)
HMAC key rotation
The signing key for v1 tokens is rotatable from the console. Rotation keeps a current + previous overlap: newly minted tokens use the current key, while tokens signed with the previous key still verify until they age out. The teardown on a Pro license lapse blanks only the dispatch maps and never the keys, so in-flight links keep resolving and a renew resumes minting with the same key.
Settings reference
Settings live in the 
parameters2 table under 
module = 'linkguard' (not 
system_settings). On save they are pushed to the engine via 
POST /api/config.
Setting
Default
Meaning
enabled
0
Master on/off for Link Guard
redirect_base_url
(console host)
Public base URL for 
/lg/ links
action_clean
redirect
Action for clean verdicts
action_suspicious
warn
Action for suspicious verdicts
action_malicious
block
Action for malicious verdicts
restore_outbound
1
Unwrap tokens on outbound replies/forwards
token_ttl_days
14
How long a rewritten link stays valid
max_inline_url
4000
Fallback-only length bound for v1 stateless tokens
rate_limit_per_min
120
Per-client-IP rate limit on 
/lg/
flag_cloud_storage
1
Flag abused cloud-storage/redirector hosts as suspicious (warn)
follow_redirects
0
Follow 
30x redirect chains at click time (guarded outbound fetch)
follow_max_hops
5
Max hops to follow when 
follow_redirects is on
cache_ttl_clean_hours
24
Verdict cache lifetime — clean
cache_ttl_suspicious_hours
6
Verdict cache lifetime — suspicious
cache_ttl_malicious_hours
168
Verdict cache lifetime — malicious
feed_urlhaus_enabled
1
Pull the URLhaus blocklist feed
feed_openphish_enabled
1
Pull the OpenPhish blocklist feed
feed_refresh_minutes
60
Feed refresh interval
gsb_enabled / 
gsb_api_key
0 / —
Google Safe Browsing lookups (optional key)
vt_enabled / 
vt_api_key
0 / —
VirusTotal lookups (optional key)
clicks_retention_days
90
Click-log retention for reporting
Two additional console-managed lists drive the verdict pipeline:
Protected recipient domains (
linkguard_domains) — which recipient domains have their inbound links rewritten. A 
_default catch-all entry protects all domains.
URL allow / block rules (
linkguard_url_rules) — operator allow/block patterns that take precedence over feeds and heuristics (layers 1–2 above).
Abused / redirector hosts (
linkguard_abused_hosts) — the seeded, operator-managed warn list for the abused-host heuristic (layer 6); see Redirect detection.
Reputation feeds and optional API lookups
URLhaus and OpenPhish are pulled on the 
feed_refresh_minutes interval into the container's SQLite store and matched as exact URL-hash lookups (a phishing URL on a shared host blocks only that URL, not the whole host).
Google Safe Browsing and VirusTotal are off by default; enable each and supply an API key to add a string-reputation layer. Results are cached per the cache-TTL settings to limit API calls.
Setting up VirusTotal and Google Safe Browsing
Both are optional — Link Guard works without them (admin lists + feeds + heuristics + redirect-following). They add a malicious-verdict layer that checks the URL as a string against the provider (the target is never fetched). Each is enabled the moment you save its key (no Save & Reload), and a malicious result will escalate even a warn-tier link to a hard block.
Where to enter keys: Link Guard page → Reputation sources → the provider's Edit key button. Entering a key auto-enables the provider; clearing it disables and wipes the stored key.
VirusTotal (recommended):
Create a free account at virustotal.com and sign in.
Open your profile menu → API key, and copy it.
Paste it into Link Guard → VirusTotal → Edit key.
Quota: the free Public API is rate-limited (≈ 4 lookups/minute, 500/day, ~15.5k/month) and is for non-commercial use; high-volume or commercial sites need a paid Premium API key. The verdict cache (and the per-tier cache-TTL settings) is what keeps you under quota — which is why Clear verdict cache warns before re-querying. A verdict counts as malicious only when ≥ 2 vendors flag the URL.
Google Safe Browsing:
In the Google Cloud Console, create (or pick) a project.
APIs & Services → Library → search "Safe Browsing API" → Enable.
APIs & Services → Credentials → Create credentials → API key, and copy it.
Paste it into Link Guard → Google Safe Browsing → Edit key.
⚠️ Commercial-use caveat: the Safe Browsing Lookup API v4 that Link Guard uses is free but designated non-commercial only by Google, and is deprecated in favour of the paid Web Risk API for commercial use. Review Google's usage terms for your deployment. If you only want one reputation provider, VirusTotal is the simpler choice; Web Risk support may be added later.
Privacy: with either provider enabled, the clicked URL string is sent to that provider's API for the lookup. Nothing else leaves Link Guard.
Branded interstitials
The warning and block pages are served by the container (
templates.py) and carry Hermes SEG branding — an inline logo, a "Hermes SEG Link Guard" header, and a footer link to hermesseg.io — rather than a generic browser error. The warning page shows the resolved host so a user can judge a shortened or suspicious link before proceeding.
Reporting and diagnostics
The admin page includes:
Check a URL — enter any URL to see the live verdict, which pipeline layer decided it, and the resolved host. This is side-effect-free (
verdict.resolve(cache_write=False)) so it does not pollute the cache.
Recent activity — a table of recent clicks (Time, Recipient domain, Host, Verdict, Decided by, Detail, Action). The Host column is the link's original host; the Detail column carries the verdict detail per click — e.g. 
redirect chain -> evil.example, 
Safe Browsing: SOCIAL_ENGINEERING, 
open-redirect -> … — so a follow-decided click shows its final/destination host right in the table (host-only, never the full URL). The 
redirect-follow container log below additionally captures follows that came back clean (which still produce a click row, now with the chain in Detail).
Redirect-follow log — when 
follow_redirects is on, every actual follow is logged (hosts only, never full URLs):
docker logs hermes_linkguard | grep "redirect-follow:"
# redirect-follow: bit.ly -> evil.example (1 hop) verdict=malicious
This shows what got followed, the hop chain, the hop count, and the resulting verdict — including follows that came back clean (which leave no row in Recent activity).
Troubleshooting commands — a collapsible card of 
docker exec one-liners for inspecting the scope map, store, and feeds.
Clear verdict cache — a button at the top of the page that flushes the entire verdict cache (every source, including GSB / VirusTotal and feed-derived results) via the mgmt endpoint 
POST /api/cache-clear, forcing a complete re-evaluation on next click. Its confirm modal warns that this re-queries Google Safe Browsing / VirusTotal, consuming additional lookups against your API quota. You rarely need it: a config edit already auto-purges the config-dependent cache (see below).
Verdict cache invalidation
The engine caches each URL's verdict (per-source TTL — clean 24h, suspicious 6h, malicious 168h) to avoid re-running heuristics / re-hitting APIs on every click. Two things keep it from going stale against admin edits:
Automatic, scoped: when 
config.json reloads after any console change (remove an abused host, add/remove a URL rule, toggle follow), 
config.py drops the config-dependent cached verdicts (
admin / 
heuristic / 
redirect / 
none) so the edit takes effect on the next click — no TTL wait, no button. Feed (
local) and external (
gsb/
vt) caches are kept (they don't depend on console config). This is why removing a host stops the warning immediately.
Manual, full: the Clear verdict cache button above flushes everything, for the rare "force a complete re-check / I think an external result is stale" case (it re-queries GSB/VirusTotal, so it's confirm-gated).
Deployment — in-stack or separate host
Link Guard runs inside the Hermes SEG stack (the default; 
hermes_linkguard service on the compose network) or on a separate host for isolation and scale. The nginx 
/lg/ location lives in the vhost template; it is delivered by regenerating the per-domain vhosts from the template (via the headless 
schedule/regen_nginx_config.cfm), not by hand-editing a generated vhost.
When Link Guard runs off-box, the milter cannot reach the shared 
url_map.db, so it mints v1 stateless tokens — the same fallback path described above. Mail flow is never affected by the container's location or availability.
Failure semantics
Link Guard is graceful-degradation by design, consistent with the rest of the body milter:
Link Guard container down / unreachable → the milter falls back to v1 stateless tokens (rewrite still happens); already-delivered links cannot be resolved until the container returns, but mail keeps flowing (
milter_default_action = accept).
Shared store unavailable → milter mints v1 tokens; mail flow never depends on the store.
Scope map empty (e.g. after a Pro license-lapse teardown) → no inbound links are rewritten, but mail flows unmodified. Re-enabling / re-saving Link Guard repopulates the map.
External API (GSB/VT) error or timeout → that layer is skipped; the verdict falls through to the next layer (worst case 
clean).
In every failure case the worst outcome is a missed rewrite or a fall-through verdict — never lost mail.
Files and data locations
Path
Container
Contents
/etc/hermes/body_milter/linkguard/linkguard_by_recipient_domain
body_milter
Scope map: protected recipient domains (
_default = all)
/var/lib/linkguard/url_map.db
body_milter (writer) / linkguard (reader)
v2 token id → original URL store, on the shared 
linkguard_data volume
/opt/linkguard/app/
linkguard
Engine code (server, verdict, feeds, store, token, templates)
Operational SQLite store
linkguard
Verdict cache, feed entries, click log
The scope map is mtime-watched by the milter and reloaded on the next message when it changes — no explicit milter reload step is needed after a console save.
Security properties (summary)
SSRF-fenced — steps 1–7 never fetch the target; the optional redirect-follow (step 8) is the only fetch, and every hop is restricted to http/https + standard ports + hosts that resolve to public IPs only.
Hard blocks are unbypassable — 
/lg/proceed re-authorizes server-side.
Hardened port split — public click surface cannot push config or read keys.
Rate-limited public surface (
rate_limit_per_min, per client IP).
Signature-safe — inbound S/MIME, PGP, and upstream-DKIM-signed mail is skipped, never re-bodied.
Mail-flow-safe — the container being down, off-box, or torn down on a license lapse never blocks delivery.

Organizational Signatures
Organizational Signatures
Pro Edition feature. Maps to Email Policies > Org Signatures (
view_org_signatures.cfm, 
edit_org_signature.cfm, 
org_signature_delete.cfm).
Hermes attaches a centrally-managed signature to outbound mail at the gateway. Admins design the signature once per domain (and optionally per department); every user on that domain gets a personalized version of it on every outbound message — no per-user setup required.
Two signature types, one pipeline
Hermes ships two distinct signature concepts that run through the same body milter and the same resolver:
Type
Tier
Owner
Storage
Per-domain control
Personal Signature
Community + Pro
The user (in 
/users/2/view_signature.cfm)
user_signatures table, one row per user
Toggled via 
domains.allow_user_signatures
Organizational Signature
Pro only
The admin (in 
Email Policies > Org Signatures)
org_signatures table, one row per 
(domain_id, department_label)
One default per domain + optional per-department variants
The milter never decides which one to apply at message time. The CFML resolver picks a winner per mailbox at admin-action time and writes a precomputed 
sender → option map; the milter just looks up the option and applies whatever it finds.
Department names — single source of truth
Departments are defined once on the mailbox edit form (Email Server > Mailboxes > Edit Options > Personal Information > Department), as free-text values typed by the admin. There is no separate "Departments" table; a department exists as soon as one mailbox is in it.
The Org Sig form's Department field is a strict dropdown sourced from the distinct 
mailboxes.department values for the selected domain. This means:
You cannot create an Org Sig for a dept that has no mailboxes — the dept won't appear in the dropdown.
The dept name on both sides is guaranteed to match exactly. No typo-class drift.
Workflow: assign at least one mailbox to the new dept first, then come back and create the Org Sig targeting it.
Changing the domain in the Org Sig form repopulates the dropdown with that domain's depts via JavaScript (no AJAX round-trip; the per-domain map is dumped into a JS const at page load).
The mailbox edit form's Department field is a free-text input with a 
<datalist> typeahead showing the same per-domain dept list. Admins can pick an existing dept (auto-completes) or type a brand-new dept name (which then appears in the dropdown next time).
If you edit an existing Org Sig whose 
department_label no longer matches any current mailbox (the dept was renamed elsewhere, or all mailboxes in it were reassigned), the orphan value is preserved in the dropdown with a 
(no mailboxes) suffix so you can still see and edit/delete the row instead of silently losing the value.
The resolver at send time does a case-insensitive trimmed match against 
mailboxes.department, so casing or whitespace drift across edits is forgiving even in the rare cases the dropdown is bypassed (e.g. direct SQL changes).
Resolution order
For every enabled mailbox, the resolver walks this priority chain top-down and stops at the first match:
1. Personal Signature
   └─ if domains.allow_user_signatures = 1
      AND user_signatures has an enabled, non-empty row for this mailbox
   ─> wins. option = user_<sanitized_email>
2. Department Organizational Signature
   └─ else if mailboxes.department is non-empty
      AND org_signatures has enabled = 1 row matching
          (domain_id, department_label)
   ─> wins. option = org_<row_id>
3. Domain Default Organizational Signature
   └─ else if org_signatures has enabled = 1 row matching
          (domain_id, department_label IS NULL)
   ─> wins. option = org_<row_id>
4. None
   └─ no map entry; the milter applies no signature to this sender's mail.
The chain is per-mailbox, not per-message. The resolver runs at admin-action time (see Triggers), serializes the winning option for every mailbox into one map file, and the milter consults that map at send time. There is no per-message DB query and no per-message resolution logic in the milter.
Pipeline placement
Same chain as Disclaimers (#214) — see disclaimers.md for the full Postfix / OpenDKIM / Ciphermail picture. The summarized order:
External MTA / MUA submission
        │
        ▼
Postfix smtpd
   ├─ smtpd_milters chain (in order):
   │   1. OpenDKIM            (signs/verifies)
   │   2. OpenDMARC           (DMARC policy)
   │   3. hermes_body_milter  (THIS — signatures, then disclaimers)
   ▼
Amavis  →  Ciphermail  →  Postfix :10026 (DKIM sign) → external
Inside the body milter, SignatureModifier runs before DisclaimerModifier, so the layered output on the outbound message is:
[user body]
[signature]      ← Personal or Organizational, picked by resolver
[disclaimer]     ← if a disclaimer is configured for this sender
OpenDKIM signs at the 
:10026 re-injection — after both modifiers — so Hermes' own DKIM signature always covers the recipient's view of the message (with signature and disclaimer baked in).
Templates
Phase 2A ships six bundled templates under 
/admin/2/inc/org_signature_templates/:
Template key
Layout
modern_card
Logo left, accent bar, contact stack right
two_column_pro
Left contact, right org block + CTA button
with_social_bar
Vertical contact + horizontal social-icon row
banner_with_logo
Full-width banner with logo top, contact below
promo_footer
Contact + bottom promotional image with link
compact_text
Minimal text-only, no images, no styling
Each template is a 
.cfm file that declares its field schema (text / email / url / color / checkbox / image fields with optional 
showIf gating) and renders pixel-perfect HTML when the renderer is invoked. Admins fill in the form on 
edit_org_signature.cfm; the gallery thumbnail + live sandboxed-iframe preview show the result before save.
All the auto-filled fields — Name, Title, Phone, Mobile, Email (
{{user.*}} from the mailbox row) plus Website and Address (
{{org.*}} from the domain row) — are collapsed under an "Override auto-filled fields" toggle by default. The admin doesn't see or edit them in the common case; the placeholders flow through to the rendered HTML unchanged and the milter fills in the recipient's data at send time. Toggling the override on exposes the fields for the rare cases that need literal text instead of substitution (shared mailboxes without personal info, seasonal URL overrides, etc.).
The genuinely admin-supplied fields stay always visible — Logo, accent color, show/hide toggles for each line, CTA text and URL, social URLs, disclaimer text, and any template-specific extras (banner height, promo image, etc.). These are the admin's actual editing surface.
Net workflow: pick a template, upload a logo, set the accent color, save. Done. Every mailbox on the domain gets a fully personalized signature on its next outbound message without any per-user form input.
Templates are version-controlled in the repo, not in the database. To add a new template, drop a new 
.cfm file in 
org_signature_templates/, add its key to the registry in 
inc/org_signature_template_loader.cfm, and produce a 240×140 thumbnail PNG. No schema migration needed.
Placeholder substitution at send time
The signature HTML stored in 
org_signatures.rendered_html (and on disk in 
body.html) keeps 
{{namespace.field}} tokens literal. The body milter substitutes them per recipient at message-send time against the sender's row in 
sender_data.json.
Available placeholders (Phase 2B):
Token
Source column
{{user.first_name}}
mailboxes.first_name
{{user.last_name}}
mailboxes.last_name
{{user.title}}
mailboxes.title
{{user.phone}}
mailboxes.phone
{{user.mobile}}
mailboxes.mobile
{{user.department}}
mailboxes.department
{{user.email}}
mailboxes.username
{{org.name}}
domains.org_name
{{org.phone}}
domains.org_phone
{{org.address}}
domains.org_address
{{org.website}}
domains.org_website
{{org.logo_url}}
domains.org_logo_path (raw URL — not cid: extracted)
{{dept.name}}
mailboxes.department
Tokens whose corresponding field is empty resolve to empty string, not literal 
{{...}}. So if 
mailboxes.title is blank for a given user, the 
{{user.title}} token disappears cleanly from delivered mail. Unknown namespaces (anything outside 
user, 
org, 
dept) are also substituted to empty.
The substitution is a single regex pass on the body html and body text inside 
SignatureModifier.modify() — well under a millisecond per message. The map and 
sender_data.json both live in process memory, refreshed only when their file mtime changes.
No DB connection from the milter, ever. All resolution and substitution data is precomputed by CFML and dropped on disk; the milter consumes the file artifacts.
Triggers — when the resolver re-runs
Both 
signature_by_sender and 
sender_data.json are rewritten in full by 
inc/signature_regen_map.cfm on every event that could affect a winner or a substitution value:
Event
Why it matters
Admin saves an Org Sig
New / edited row may win for senders that previously had no match or a lower-tier match
Admin deletes an Org Sig
Losers fall back to the next tier (or none)
Admin edits a domain (
allow_user_signatures or 
org_* columns)
Per-domain toggle flips the Personal-vs-Org winner; org_* values feed 
{{org.*}} substitution
Admin edits a mailbox (Pro fields: first_name, title, dept, etc.)
{{user.*}} and 
{{dept.name}} substitution data changes; a department change can flip Default → Department winner
Admin adds a mailbox
New sender enters resolution and may pick up a domain-default Org Sig
Admin deletes a mailbox
Sender must drop from the map
User saves their Personal Signature
May now win over the previously-resolved Org Sig (or vice versa if disabling)
Each trigger runs the same shared resolver. Full rebuild every time, not incremental. With low-thousands of mailboxes the rebuild is well under a second, and the simplicity rules out drift bugs ("did we forget to update X for sender Y" can't happen).
The body milter mtime-watches both files and reloads in process memory on the next message after the file changes. No SIGHUP, no IPC, no container restart.
Files generated on save
The CFML resolver writes:
/etc/hermes/body_milter/signatures/signature_by_sender   sender → option map
/etc/hermes/body_milter/signatures/sender_data.json       sender → {{token}} dict
/etc/hermes/body_milter/signatures/files/<option>/
    body.txt         plain-text signature (auto-derived from html)
    body.html        html signature with cid: refs (placeholders intact)
    images/
        1.png        per-option inline images (#230 pattern)
        2.jpg
        ...
Where 
<option> is 
user_<sanitized_email> (Personal Sig) or 
org_<row_id> (Org Sig). 
<sanitized_email> is 
bob.smith@example.com → 
bob_smith_at_example_com (
@ → 
_at_, non-alphanumerics → 
_).
signature_by_sender example:
alice@example.com   org_42
bob@example.com     user_bob_at_example_com
carol@example.com   org_43
sender_data.json example (post-Lucee uppercasing — the milter normalizes to lowercase on load):
{
  "alice@example.com": {
    "USER.FIRST_NAME": "Alice",
    "USER.TITLE": "Sales Manager",
    "ORG.NAME": "Acme",
    "ORG.PHONE": "555-0100",
    "DEPT.NAME": "Sales"
  }
}
The 
files/<option>/ dir is wiped before re-render on every save of that row, so deleted images and renamed scope keys never leave stale binaries behind.
Inline images (cid: extraction)
Same pattern as Disclaimers (#230) — see disclaimers.md for the MIME / multipart-related details.
For Org Signatures, the cid: namespace is 
signature_org_<row_id>_img_<N> (Personal Signatures use 
signature_user_<sanitized_email>_img_<N>). Both share the milter regex 
cid:(signature_[\w.-]+_img_\d+), so cid: refs from either signature type are queued for inline attachment alongside any cid: refs from a domain disclaimer on the same message — no namespace collisions.
Per-template image fields use the same data: → cid: pipeline as user-pasted Personal Sig images. At admin-save time:
Admin uploads the file in the form (or pastes a URL — both are handled).
Browser converts the file to a 
data:image/...;base64,... URI via 
FileReader, capped at 1 MB per image.
CFML renders the template; the resulting HTML carries the data: URI inline.
inc/org_signature_write_files.cfm extracts each 
data: URI, decodes the base64 into a binary file under 
images/, and rewrites the html to reference 
<img src="cid:signature_org_<id>_img_<N>">.
At message-send time the milter walks the cid: refs, attaches each image as an 
image/<format> MIME part with 
Content-ID and 
Content-Disposition: inline, and wraps the message as 
multipart/related.
{{org.logo_url}} is not cid: extracted — it's a raw URL substituted into the html as-is. Use it for hosted-elsewhere logos (your CDN, your website). Use the per-template image field for cid:-attached inline logos when you want them to render even in MUAs that block external images.
Behavior with S/MIME, PGP, DKIM-signed mail
Identical to Disclaimers — same skip rules in the same Modifier base class. Pre-signed envelopes, PGP inline, and pre-existing DKIM-Signature headers all cause the body milter to leave the message untouched.
See disclaimers.md "Behavior with S/MIME, PGP, and DKIM-signed mail" for the table of patterns and the operational consequences.
Reply-chain handling
No dedup — every outbound gets a fresh signature, including replies inside a long thread. Same industry-norm pattern Disclaimers uses; same rationale (compliance, self-contained messages, predictability). See disclaimers.md "Reply-chain handling".
Failure semantics
Same graceful-degradation contract as Disclaimers (
milter_default_action = accept). If the milter container is down, if the map file is unreadable, if the modifier raises an exception, if substitution blows up — mail flows unmodified. Worst case is a missed signature; mail never gets dropped.
See disclaimers.md "Failure semantics".
Disabled rows
org_signatures.enabled = 0 causes the resolver to skip the row entirely:
The on-disk 
org_<id>/ dir is wiped clean
No mailboxes resolve to that option
Mailboxes that previously resolved to that option fall back to the next tier (department → default → none)
Re-enabling rebuilds the on-disk files and re-points the affected mailboxes' map entries on the next regen.
Interaction with 
domains.allow_user_signatures
This per-domain toggle is the single switch that controls whether Personal Signatures can win over Organizational Signatures.
allow_user_signatures
Personal Sig present?
Result
0
yes
Personal Sig ignored; resolver falls to Department / Default Org Sig
0
no
Resolver falls to Department / Default Org Sig
1
yes
Personal Sig wins (top of resolution chain)
1
no
Resolver falls to Department / Default Org Sig
Toggle this off when you need to lock everyone into Organizational Signatures for branding/compliance — useful when a marketing or legal team wants centrally-controlled output and doesn't want individual users overriding it.
A previously-saved Personal Signature is not deleted when the toggle goes off — it just stops being resolved. Toggling back on re-activates it on the next regen.
Pro license behavior
The Org Signatures admin page is gated by 
session.edition EQ "Pro":
Pro license valid: Full UI access; admins can create / edit / delete / enable / disable Org Signatures and toggle 
allow_user_signatures.
Pro license missing or expired: Sidebar entry shows a 
PRO badge; the list and edit pages reject the load with an upsell flash. The save action handler ALSO rejects on the server side (defense in depth — UI-level gating alone isn't a security control).
Existing Org Signatures on a downgrade: Rows persist in the database. The resolver still runs and the milter continues applying them at send time. Personal Signatures continue working as well. The downgrade is a UI restriction, not a runtime feature kill.
This is the same "feature stays on, admin UI locks" pattern as Disclaimers and other Pro features. If a customer wants the Pro feature actually disabled at runtime on downgrade, the path is to delete the rows or set them all to 
enabled = 0.
Why a separate milter and not an amavis hook
Same reasoning as Disclaimers (#214 Phase 3). amavisd-new 2.13's 
before_send hook silently desynchronizes amavis's internal MIME state during in-place body modification, which can drop mail. The body milter approach moves body modification out of amavis entirely; amavis is fully decoupled.
See disclaimers.md "Why a separate milter and not an amavis hook".