Make the Signed-Commit Verification Tech-Agnostic

[ptr727]

Why

Closes the root cause behind #706: an agent on a Windows host concluded commit
signing was broken and stopped work, based on git config --get commit.gpgsign && ssh-add -L. That check tests one specific delivery path (an agent holding
the key). gpg.format=ssh can sign straight from a key file with no
ssh-agent running at all -- this host's actual setup -- so the probe failed
while signing itself worked cleanly the whole time.

A follow-up comment on the same issue chased a second red herring: mixed
GPG/SSH signatures in develop's history, read as evidence of a
still-misconfigured fleet host. Investigating it here found no such host --
every GPG-signed commit on develop is a squash-merge, committed and signed
by GitHub <noreply@github.com> itself, server-side, regardless of what the
PR author signed with locally. That's structural on every fleet repo, not
drift.

What

In .agents/skills/git-commit-conventions/SKILL.md (.claude-plugin/
regenerated via scripts/build_dist.py):

• Bans agent-liveness probes (ssh-add -L, a gpg-agent check) as an
  invalid signing test, and states why.
• Replaces it with a real scratch commit read back via git's own %G?
  verdict character -- tech-agnostic across SSH agent-backed, SSH key-file,
  GPG agent-backed, and GPG keyring signing, and immune to git version/locale
  text differences that a "Good" grep would not be.
• The same scratch commit's %ae now backs the identity check too, so one
  probe verifies both signing and identity instead of trusting
  git config values that don't prove what lands on the commit object.
• Documents the GitHub-committer/GPG-at-every-squash-merge pattern so an
  agent checks commit.committer.name before treating a differing signature
  type in history as a clue worth chasing.

Verified the new probe directly on this host (sig=G email=ptr727@users.noreply.github.com) before committing with it.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR updates the git-commit-conventions skill documentation to make commit-signing verification tech-agnostic by replacing agent-liveness probes with a scratch-commit check that reads Git’s %G? signature verdict and %ae identity output end-to-end.

Changes:

• Replace ssh-add -L / agent-liveness probing guidance with a scratch-commit verification flow using git log --format='sig=%G? email=%ae'.
• Document the expected GitHub squash-merge committer/signature pattern to avoid misdiagnosing mixed signature history.
• Regenerate the distributed .claude-plugin skill output and update the fleet-skills source digest.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
.agents/skills/git-commit-conventions/SKILL.md	Updates the canonical skill text to use a scratch-commit signing + identity verification and clarifies mixed-signature history expectations.
.claude-plugin/fleet-skills/skills/git-commit-conventions/SKILL.md	Updates the generated/distributed copy of the skill.
.claude-plugin/fleet-skills/.source-digest	Updates the digest for the regenerated fleet-skills output.

Suppressed comments (2)

.agents/skills/git-commit-conventions/SKILL.md:79

• This paragraph uses an em dash ("squash-merge — GitHub"), which violates the repo's ASCII-only rule for agent-authored text (spec/project-types.json:178). Please replace it with ASCII punctuation (for example, a spaced hyphen).

  `git log --pretty='%G? %GK'` shows two distinct shapes, not two health states: a commit committed
  by the PR's own author carries that host's own signature type, while a commit committed by
  `GitHub <noreply@github.com>` is a squash-merge — GitHub creates and signs that commit itself,
  server-side, with GitHub's own GPG key, regardless of what the PR author signed with locally.

.agents/skills/git-commit-conventions/SKILL.md:99

• This sentence uses an em dash ("end-to-end — read"), which violates the repo's ASCII-only rule for agent-authored text (spec/project-types.json:178). Replace it with ASCII punctuation (for example, a spaced hyphen).

or `ID+username@users.noreply.github.com` form. **Verify it, do not set it**: the scratch commit
from the signing check above already proves this end-to-end — read its `email=` output rather
than trusting `git config --get user.email` alone, since a global config value doesn't prove what
actually lands on a commit object. Match it against that address before committing, rather than

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

Suppressed comments (6)

STANDUP.md:61

• This paragraph also references git-commit-conventions.md, which is not a file in this repository. Update it to the actual path so the pointer is actionable.

The scratch commit exercises the whole signing pipeline rather than one delivery path, since `ssh-add -L` or `gpg --list-secret-keys` only prove an agent holds a key and say nothing about a host that signs straight from a key file with no agent running at all, a live and correctly configured case documented in `git-commit-conventions.md` "Signing, verified not configured", which [GOVERNANCE.md "Git and Commit Rules"][governance-git-and-commit-rules] points to. Signing is **SSH or GPG**, so this judges the configured format by its actual result (`sig=G`), never by which delivery path produced it. A missing `--global` value, `sig` not reading `G`, or `email` not matching the noreply address is a **host** misconfiguration to surface to the maintainer ([`docs/host-setup.md`][host-setup] is the setup procedure), not something to patch per repo. Patching it locally hides a broken host that then produces wrong identities in every other repo on that machine.

.agents/skills/git-commit-conventions/SKILL.md:57

• This section correctly states that ssh-add -L is not a reliable signing verification probe, but docs/host-setup.md still presents ssh-add -L as part of the host verification checklist. That can reintroduce the same false-negative behavior described in #706 unless the host-setup doc is updated to match this guidance.

  agent-backed or straight from a keyring. **Probing agent liveness (`ssh-add -L`, a `gpg-agent`
  check) is not a valid test and must not be used.** It tests one specific delivery path, not
  whether `git commit -S` succeeds, and a host that signs straight from a key file fails that
  probe while signing correctly.

STANDUP.md:37

• git-commit-conventions.md is referenced here, but that file does not exist in this repo. Point this at the actual skill file so readers can find the rationale being referenced.

This issue also appears on line 61 of the same file.

# prove signing works with a live scratch commit, not an agent-liveness probe (ssh-add -L,
# gpg --list-secret-keys): a host that signs straight from a key file with no agent running
# passes cleanly and fails that probe. See git-commit-conventions.md "Signing, verified not
# configured" for why.

STANDUP.md:40

• The scratch-commit probe ends with rm -rf "$d" on its own line, which makes the snippet's exit status always reflect the cleanup, not whether git init / git commit -S / git log succeeded. Wrapping cleanup in a subshell trap keeps the temp dir cleanup while preserving the failure status for scripted/agent execution.

d=$(mktemp -d) && git init -q "$d" && git -C "$d" commit -S --allow-empty -q -m check \
  && git -C "$d" log -1 --format='sig=%G? email=%ae'
rm -rf "$d"

.agents/skills/git-commit-conventions/SKILL.md:66

• The scratch-commit snippet does cleanup as its final command (rm -rf "$d"), which means the overall exit status reflects the cleanup rather than whether the signing/git log probe succeeded. If an agent runs this in a script, failures can be masked. Wrap the body in a subshell with a cleanup trap so the probe's exit status is preserved while still cleaning up the temp directory.

  d=$(mktemp -d) && git init -q "$d" \
    && git -C "$d" commit -S --allow-empty -q -m check \
    && git -C "$d" log -1 --format='sig=%G? email=%ae'
  rm -rf "$d"

.claude-plugin/fleet-skills/skills/git-commit-conventions/SKILL.md:66

• Same as the source skill: the scratch-commit snippet ends with cleanup as the last command, which can hide failures if the snippet is run in a script. Use a subshell trap so the probe's exit status is preserved while still cleaning up the temp directory.

  d=$(mktemp -d) && git init -q "$d" \
    && git -C "$d" commit -S --allow-empty -q -m check \
    && git -C "$d" log -1 --format='sig=%G? email=%ae'
  rm -rf "$d"

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Suppressed comments (11)

docs/host-setup.md:269

• The PowerShell probe output only prints the author email (%ae). Include the committer email (%ce) as well so the identity verification is end-to-end and consistent with GitHub's committer-based verification.

$d = Join-Path $env:TEMP ([guid]::NewGuid())
git init -q $d
git -C $d commit -S --allow-empty -q -m check
git -C $d log -1 --format='sig=%G? email=%ae'
Remove-Item -Recurse -Force $d

.agents/skills/git-commit-conventions/SKILL.md:107

• If the probe output includes both author and committer emails, the surrounding text should tell readers to verify both values (not just email=) when checking the noreply identity.

or `ID+username@users.noreply.github.com` form. **Verify it, do not set it**: the scratch commit
from the signing check above already proves this end-to-end. Read its `email=` output rather
than trusting `git config --get user.email` alone, since a global config value doesn't prove what
actually lands on a commit object. Match it against that address before committing, rather than

.claude-plugin/fleet-skills/skills/git-commit-conventions/SKILL.md:77

• The PowerShell probe prints only the author email (%ae). Include the committer email (%ce) too so the identity check cannot miss author/committer divergence.

  $d = Join-Path $env:TEMP ([guid]::NewGuid())
  git init -q $d
  git -C $d commit -S --allow-empty -q -m check
  git -C $d log -1 --format='sig=%G? email=%ae'
  Remove-Item -Recurse -Force $d

**.claude-plugin/fleet-skills/skills/git-commit-conventions/SKILL.md:107**
* If the probe output includes both author and committer emails, the identity section should instruct readers to verify both values (not just `email=`).

or ID+username@users.noreply.github.com form. Verify it, do not set it: the scratch commit
from the signing check above already proves this end-to-end. Read its email= output rather
than trusting git config --get user.email alone, since a global config value doesn't prove what
actually lands on a commit object. Match it against that address before committing, rather than

**docs/host-setup.md:250**
* After adding committer_email to the probe output, this requirement should check both the author and committer emails, not just %ae.

sig must read G and email must match the noreply address from the config line above.

**.agents/skills/git-commit-conventions/SKILL.md:77**
* The PowerShell probe prints only the author email (%ae). Include the committer email (%ce) too so the identity verification cannot miss author/committer divergence.

$d = Join-Path $env:TEMP ([guid]::NewGuid())
git init -q $d
git -C $d commit -S --allow-empty -q -m check
git -C $d log -1 --format='sig=%G? email=%ae'
Remove-Item -Recurse -Force $d

docs/host-setup.md:245

• The scratch commit output only prints the author email (%ae). Since the guidance asks readers to verify the noreply identity, also print the committer email (%ce) so the check can catch cases where author/committer diverge.

This issue also appears in the following locations of the same file:

• line 250
• line 265

  git init -q "$d" \
    && git -C "$d" commit -S --allow-empty -q -m check \
    && git -C "$d" log -1 --format='sig=%G? email=%ae'
)

STANDUP.md:42

• The scratch commit output only prints the author email (%ae), but the surrounding guidance emphasizes that committer identity is what GitHub verifies and can diverge from the author (e.g., rebase/amend). Include %ce in the probe output so the pre-init check can validate both identities end-to-end.

    && git -C "$d" commit -S --allow-empty -q -m check \
    && git -C "$d" log -1 --format='sig=%G? email=%ae'

docs/host-setup.md:255

• This paragraph is hard-wrapped and splits the link text across a newline, which is inconsistent with the surrounding one-paragraph-per-line style and makes the link harder to maintain. Prefer a single-line paragraph and keep the link text unbroken.

`ssh-add -L` (or a `gpg --list-secret-keys` equivalent) is not a substitute: it only proves an
agent holds a key, and a host that signs straight from a key file with no agent running passes
this scratch commit while failing that probe, per [GOVERNANCE.md "Git and Commit
Rules"][governance-git-and-commit-rules]. If signing fails locally, the devcontainer will fail
too, so fix here first.

.agents/skills/git-commit-conventions/SKILL.md:67

• The scratch commit probe prints only the author email (%ae). To make this a true end-to-end identity check (and to match the later guidance about committer importance), include %ce in the output as well.

This issue also appears in the following locations of the same file:

• line 72
• line 104

  d=$(mktemp -d) && (
    trap 'rm -rf "$d"' EXIT
    git init -q "$d" \
      && git -C "$d" commit -S --allow-empty -q -m check \
      && git -C "$d" log -1 --format='sig=%G? email=%ae'
  )

.claude-plugin/fleet-skills/skills/git-commit-conventions/SKILL.md:67

• The scratch commit probe prints only the author email (%ae). Include %ce so this can actually validate both identities end-to-end, consistent with the later committer-focused guidance.

This issue also appears in the following locations of the same file:

• line 72
• line 104

  d=$(mktemp -d) && (
    trap 'rm -rf "$d"' EXIT
    git init -q "$d" \
      && git -C "$d" commit -S --allow-empty -q -m check \
      && git -C "$d" log -1 --format='sig=%G? email=%ae'
  )

[ptr727]

Answering the suppressed (low-confidence) findings across all three rounds so far, quoting each and giving a verdict:

Round on 633bf1a (2 findings)

• .agents/skills/git-commit-conventions/SKILL.md:79 — em dash in "squash-merge — GitHub"
• .agents/skills/git-commit-conventions/SKILL.md:99 — em dash in "end-to-end — read"

Fixed in 8eb9050: both rewritten to ASCII punctuation, along with four other em dashes the visible review round already caught.

Round on 8eb9050 (6 findings)

• STANDUP.md:61 and STANDUP.md:37 — both reference the non-existent filename git-commit-conventions.md
• .agents/skills/git-commit-conventions/SKILL.md:57 — docs/host-setup.md still ran ssh-add -L as part of host verification, reintroducing the exact false-negative Fix the Misleading ssh-add Signing Check in git-commit-conventions #706 was filed over
• STANDUP.md:40, .agents/skills/git-commit-conventions/SKILL.md:66, and the generated copy at :66 — the scratch probe's trailing rm -rf "$d" ran after the && chain, so the snippet's exit status always reflected cleanup rather than the signing check

Fixed in 1d705d4: the two broken references now point at the real path (.agents/skills/git-commit-conventions/SKILL.md, added as a link target in STANDUP.md); docs/host-setup.md's "Verify Host Setup" section now runs the same scratch-commit probe instead of ssh-add -L plus a hardcoded -c gpg.format=ssh (POSIX and PowerShell); all three probe copies wrap the git chain in a subshell with an EXIT trap, verified locally to preserve the probe's own exit status (sig=G, exit=0, temp dir removed).

Round on 1d705d4 (11 findings)

• docs/host-setup.md:245, :250, :265, :269; .agents/skills/git-commit-conventions/SKILL.md:67, :72, :77, :104, :107; the generated copy at :67, :72, :77, :107; STANDUP.md:42 — the scratch probe's format string printed only %ae (author email), while both this skill's own Identity section and STANDUP.md's pre-existing verification (which the probe was modeled after) require checking author and committer both, since a rebase/amend/cherry-pick rewrites the committer while leaving the author untouched
• docs/host-setup.md:255 — a paragraph was hard-wrapped, inconsistent with the file's one-paragraph-per-line style

Fixed in d3fbb1c: every probe copy (git-commit-conventions.md, STANDUP.md, host-setup.md POSIX and PowerShell) now uses STANDUP.md's own established format string, sig=%G? author=%an <%ae> committer=%cn <%ce>, with surrounding prose updated to say so; the hard-wrapped paragraph is back to one line.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Suppressed comments (3)

docs/host-setup.md:264

• The PowerShell scratch-commit snippet does not quote $d. On many Windows hosts $env:TEMP (or the username path segment) can contain spaces, which will break git init -q $d / git -C $d ... by splitting the path into multiple arguments. Also, cleanup only runs if all commands succeed, leaving temp dirs behind on failure. Quote the path and wrap cleanup in a finally block.

$d = Join-Path $env:TEMP ([guid]::NewGuid())
git init -q $d
git -C $d commit -S --allow-empty -q -m check
git -C $d log -1 --format='sig=%G? author=%an <%ae> committer=%cn <%ce>'
Remove-Item -Recurse -Force $d

.claude-plugin/fleet-skills/skills/git-commit-conventions/SKILL.md:76

• Same issue as the source skill: the PowerShell snippet does not quote $d (breaks when $env:TEMP contains spaces) and cleanup only runs on success. Since this file is the distributed copy, it should reflect the corrected try/finally + quoted-path version after regeneration.

  $d = Join-Path $env:TEMP ([guid]::NewGuid())
  git init -q $d
  git -C $d commit -S --allow-empty -q -m check
  git -C $d log -1 --format='sig=%G? author=%an <%ae> committer=%cn <%ce>'
  Remove-Item -Recurse -Force $d

.agents/skills/git-commit-conventions/SKILL.md:76

• The PowerShell equivalent does not quote $d, so it will fail on Windows hosts where the temp path contains spaces. It also only removes the temp dir if every prior command succeeds. Quote the path and use try/finally so cleanup runs even when signing fails (which is exactly what this probe is meant to detect).

  $d = Join-Path $env:TEMP ([guid]::NewGuid())
  git init -q $d
  git -C $d commit -S --allow-empty -q -m check
  git -C $d log -1 --format='sig=%G? author=%an <%ae> committer=%cn <%ce>'
  Remove-Item -Recurse -Force $d

[ptr727]

Answering round 4's suppressed findings (all on d3fbb1c):

• docs/host-setup.md:264, .agents/skills/git-commit-conventions/SKILL.md:76, and the generated copy at :76 — the PowerShell probe left $d unquoted, which breaks git init/git -C on a Windows profile path containing a space, and ran Remove-Item as a trailing statement, so a failed git command skipped cleanup and left the temp dir behind.

Fixed in 9b2e941: every $d use is quoted, and the probe body is wrapped in try { ... } finally { Remove-Item ... } so cleanup runs whether the probe succeeds or fails, in all three copies.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Suppressed comments (4)

STANDUP.md:38

• mktemp -d without a template is not portable: the default macOS/BSD mktemp requires a template (or -t), so this signing probe will fail on macOS hosts even when signing is correctly configured.

d=$(mktemp -d) && (

docs/host-setup.md:241

• The POSIX host-setup probe uses mktemp -d without a template. On macOS (supported by this doc), BSD mktemp typically requires a template (or -t), so this step can fail before running the signing check.

d=$(mktemp -d) && (

.agents/skills/git-commit-conventions/SKILL.md:63

• This scratch-commit probe uses mktemp -d without a template. That works on GNU coreutils but is not portable to macOS/BSD mktemp (which expects a template or -t), so the verification step can fail on macOS even when signing works.

  d=$(mktemp -d) && (

.claude-plugin/fleet-skills/skills/git-commit-conventions/SKILL.md:63

• This distributed copy also uses mktemp -d without a template, which is not portable to macOS/BSD mktemp (template or -t required). Update the source skill and regenerate so both copies stay consistent.

  d=$(mktemp -d) && (

[ptr727]

Answering round 5's suppressed findings (all on 9b2e941):

• STANDUP.md:38, docs/host-setup.md:241, .agents/skills/git-commit-conventions/SKILL.md:63, and the generated copy at :63 — mktemp -d with no template works on GNU coreutils but errors on BSD/macOS mktemp, which requires a template argument, so the signing probe would fail on a macOS host even when signing itself is fine.

Fixed in 0553d34: switched to mktemp -d "${TMPDIR:-/tmp}/sign-check.XXXXXX" in all three POSIX copies, the explicit-template form both GNU and BSD mktemp accept; verified locally (sig=G, exit=0).

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated no new comments.

Suppressed comments (7)

docs/host-setup.md:263

• Same as the POSIX snippet above: git commit -S forces signing and can hide a missing/false commit.gpgsign. Using a plain git commit makes this an end-to-end verification that signing is enabled by configuration.

  git -C "$d" commit -S --allow-empty -q -m check

docs/host-setup.md:267

• In the PowerShell snippet, Remove-Item in finally can throw if git init fails before creating the directory, which can mask the real error. Guard the cleanup with Test-Path so failures report the underlying signing/identity problem.

} finally {
  Remove-Item -Recurse -Force "$d"
}

.agents/skills/git-commit-conventions/SKILL.md:79

• The finally cleanup can throw if git init fails before creating $d, potentially masking the real failure you want the user to see. Guard the cleanup with Test-Path so errors from signing/identity remain visible.

  } finally {
    Remove-Item -Recurse -Force "$d"
  }

.agents/skills/git-commit-conventions/SKILL.md:75

• In the PowerShell equivalent, git commit -S forces signing and can mask a missing/false commit.gpgsign. Using a plain git commit makes the probe validate the default signing configuration end-to-end.

    git -C "$d" commit -S --allow-empty -q -m check

STANDUP.md:41

• The scratch commit uses git commit -S, which bypasses the commit.gpgsign configuration you are also asking the user to verify. Using a plain git commit here better validates that the host is configured to sign commits by default (and avoids a false pass when commit.gpgsign is unset/false).

    && git -C "$d" commit -S --allow-empty -q -m check \

docs/host-setup.md:244

• This verification snippet uses git commit -S, which can pass even if commit.gpgsign is unset/false (because -S forces signing). If the intent is to prove the host is configured to sign commits by default, use a plain git commit so the check fails when signing is not enabled by configuration.

This issue also appears in the following locations of the same file:

• line 263
• line 265

    && git -C "$d" commit -S --allow-empty -q -m check \

.agents/skills/git-commit-conventions/SKILL.md:66

• The scratch-commit probe uses git commit -S, which forces signing and can still succeed even if commit.gpgsign is unset/false. Using a plain git commit better validates the default signing configuration the rest of the guidance expects.

This issue also appears in the following locations of the same file:

• line 75
• line 77

      && git -C "$d" commit -S --allow-empty -q -m check \

[ptr727]

Answering round 6's suppressed findings (all on 0553d34):

• STANDUP.md:41, docs/host-setup.md:244 (also :263), .agents/skills/git-commit-conventions/SKILL.md:66 (also :75, :77) — the probe used git commit -S, which forces a signature regardless of commit.gpgsign. A host with signing keys configured but commit.gpgsign unset or false would still pass, which is exactly the default-config gap the probe exists to catch, since every real agent commit is plain (no -S).
• docs/host-setup.md:267, .agents/skills/git-commit-conventions/SKILL.md:79 — the PowerShell finally block's Remove-Item can throw if git init never created $d, masking the real signing/identity error underneath.

Fixed in 21aec10: dropped -S in every POSIX and PowerShell copy (verified locally: a plain commit still signs correctly here, sig=G, since commit.gpgsign=true), added a sentence explaining why -S is deliberately absent, and guarded both PowerShell finally blocks with Test-Path before Remove-Item.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Suppressed comments (4)

STANDUP.md:62

• The explanation of sig=U is too narrow. In Git, %G? can also return U for SSH-signed commits when the signature is cryptographically valid but the signer is not trusted/recognized (for example, missing from gpg.ssh.allowedSignersFile / allowed_signers). The current text says U is GPG-only, which conflicts with the rest of the guidance (and with how SSH verification works).

The scratch commit exercises the whole signing pipeline rather than one delivery path, since `ssh-add -L` or `gpg --list-secret-keys` only prove an agent holds a key and say nothing about a host that signs straight from a key file with no agent running at all, a live and correctly configured case [git-commit-conventions][git-commit-conventions] documents in "Signing, verified not configured", the same rules [GOVERNANCE.md "Git and Commit Rules"][governance-git-and-commit-rules] points to. Signing is **SSH or GPG**, so this judges the configured format by its actual result (`sig=G`, or `sig=U` for a cryptographically good GPG signature whose key trust is merely undefined), never by which delivery path produced it. A missing `--global` value, `sig` not reading `G` or `U`, or either printed email not matching the noreply address is a **host** misconfiguration to surface to the maintainer ([`docs/host-setup.md`][host-setup] is the setup procedure), not something to patch per repo. Patching it locally hides a broken host that then produces wrong identities in every other repo on that machine.

docs/host-setup.md:247

• This paragraph states that sig=U is GPG-only and that SSH has no trust concept. For SSH signing, Git can still report %G? = U when the signature is cryptographically good but the signer is not trusted/recognized (for example, missing from gpg.ssh.allowedSignersFile). The current wording may mislead readers into treating U as impossible under SSH.

`sig` must read `G` (good signature) or `U` (good signature, undefined trust: GPG-only, common on a freshly generated key before its trust is set to ultimate; SSH's `allowed_signers` carries no trust concept, so this never applies there), both the `author` and `committer` email must be an actual noreply address, and both must match `user.email` from the config line above, all enforced by the snippet itself. `ssh-add -L` (or a `gpg --list-secret-keys` equivalent) is not a substitute: it only proves an agent holds a key, and a host that signs straight from a key file with no agent running passes this scratch commit while failing that probe, per [GOVERNANCE.md "Git and Commit Rules"][governance-git-and-commit-rules]. If signing fails locally, the devcontainer will fail too, so fix here first.

.agents/skills/git-commit-conventions/SKILL.md:97

• The sig=U explanation is incorrect for SSH signing. Git's %G? can return U for SSH-signed commits when the signature is valid but the signer is not trusted/recognized (for example, not listed in gpg.ssh.allowedSignersFile). The current text says SSH has no trust concept and that U never applies, which can cause agents to mis-diagnose a working signing setup as broken (or vice versa).

  `sig` must read `G` (good signature) or `U` (good signature, undefined trust: GPG-only, common
  on a freshly generated key before its trust is set to ultimate). SSH's `allowed_signers` carries
  no trust concept, so `U` never applies there. `sig` is git's own verdict char. Don't grep localized

.claude-plugin/fleet-skills/skills/git-commit-conventions/SKILL.md:97

• Same issue as the source skill: the sig=U explanation claims SSH never yields U, but Git can return %G? = U for SSH signatures when the signer is not trusted/recognized (for example, missing from gpg.ssh.allowedSignersFile). This should match the corrected wording in the source so the distributed copy is accurate.

  `sig` must read `G` (good signature) or `U` (good signature, undefined trust: GPG-only, common
  on a freshly generated key before its trust is set to ultimate). SSH's `allowed_signers` carries
  no trust concept, so `U` never applies there. `sig` is git's own verdict char. Don't grep localized

[ptr727]

Answering round 16's suppressed findings (all on 4f03107):

• STANDUP.md:62, docs/host-setup.md:247, .agents/skills/git-commit-conventions/SKILL.md:97, and the generated copy at :97 — my own round-12 claim ("SSH's allowed_signers carries no trust concept, so U never applies there") was wrong.

Verified this directly rather than take either side's word: signed a scratch commit with gpg.ssh.allowedSignersFile pointed at an empty file, and got exactly sig=U, Good "git" signature ... No principal matched.

Fixed in 63336a3: the shell logic (accept G or U) was already correct for what this probe actually needs to prove, since a missing local allowed_signers entry doesn't affect whether GitHub itself verifies the commit, only local git verify-commit output. Only the prose was wrong — corrected in all three files: U covers both a GPG key with merely undefined trust and an SSH key missing from allowed_signers, neither of which means the commit won't verify on GitHub.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Suppressed comments (1)

docs/host-setup.md:243

• The note about CRLF breaking \ line continuations is helpful, but this doc still contains at least one \-continued shell snippet (the allowed_signers creation earlier). As written, it is easy to miss that those other snippets are subject to the same copy/paste failure mode. Consider broadening this note to apply to the whole document (or rewrite the remaining \-continued snippets) so the guidance is internally consistent.

# One physical line, not backslash-joined: this file is CRLF (the repo's Markdown default),
# and a `\` continuation stops working the moment a stray `\r` lands after it.
d=$(mktemp -d "${TMPDIR:-/tmp}/sign-check.XXXXXX") && ( trap 'rm -rf "$d"' 0; email=$(git config --global --get user.email) && git init -q "$d" && git -C "$d" commit --allow-empty -q -m check && out=$(git -C "$d" log -1 --format='sig=%G? author=%an <%ae> committer=%cn <%ce>') && echo "$out" && ae=$(git -C "$d" log -1 --format='%ae') && ce=$(git -C "$d" log -1 --format='%ce') && case "$out" in sig=G\ *|sig=U\ *) true ;; *) false ;; esac && case "$email" in *@users.noreply.github.com) true ;; *) false ;; esac && [ "$ae" = "$email" ] && [ "$ce" = "$email" ] )

[ptr727]

Answering round 17's suppressed finding (on 63336a3):

• docs/host-setup.md:243 — the new CRLF-safety comment sits right next to another pre-existing \-continued snippet (the allowed_signers file creation, predating this PR) that has the exact same fragility, making the guidance internally inconsistent.

Fixed in c7fb717: collapsed that snippet to one line, same fix as the scratch probe itself.

That same push also carries a real fix from a Windows-side agent (840388d): my round-15 PowerShell change (git ... commit ... && $out = git ... log ...) doesn't parse in real PowerShell 7 (&&can't chain into an assignment statement). I couldn't verify that myself locally (nopwsh` on this host); the Windows-side session validated it on an actual Windows 11 / PowerShell 7.6.4 host and fixed it by pulling the assignment onto its own statement.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

Suppressed comments (1)

docs/host-setup.md:140

• The allowed_signers entry is built from git config user.email, which can pick up a repo-local override if this snippet is run from inside an existing repository. Since this section is configuring host-level signing and the rest of the docs now emphasize --global for identity checks, read the global value explicitly here to avoid writing a mismatched principal into ~/.config/git/allowed_signers.

echo "$(git config user.email) namespaces=\"git\" $(cat ~/.ssh/id_ed25519.pub)" >> ~/.config/git/allowed_signers

[ptr727]

Answering round 19's suppressed finding (on c7fb717):

• docs/host-setup.md:140 — the allowed_signers entry was built from a bare git config user.email, which picks up a repo-local override if this snippet is run from inside an existing repo checkout, silently writing the wrong principal into ~/.config/git/allowed_signers.

Fixed in 1d7cc89: added --global, matching every other identity read this PR touches.

docs/ssh-signing.md carries the identical line (same missing --global, same CRLF-fragile backslash continuation) but is untouched by this PR and outside what the review scans (Copilot only reviews changed files). Filed #713 to track it rather than expand this PR's scope.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.