Repository navigation
Description
CONTRIBUTING.md:45 states that every commit must carry a Signed-off-by trailer, and CONTRIBUTING.md:53 states that PRs with commits lacking sign-off will not merge.
Since squash merges can drop trailers from the merged history, I looked at the PR commits themselves, which is where the rule applies. Across the 20 most recently merged PRs:
| PRs examined | 20 |
| Commits across those PRs | 113 |
Commits carrying Signed-off-by |
8 |
| PRs with at least one signed commit | 4 (#1146, #1138, #1125, #1046) |
| PRs where every commit is signed | 3 (#1138, #1125, #1046) |
So 16 of the 20 PRs have no signed commits at all, and the 8 that are signed sit in a handful of PRs rather than following a date or an author.
I could not find a check for it inside the repository either: .github/workflows/ci.yml runs npm install, npm run build, npm run skills:check and npm test, and there is no DCO workflow or config file under .github/.
I also looked at what actually ran on those PRs, on both APIs. Across the 20 PR head commits:
- Check runs: the only names that appear are the CI test matrix (
test (ubuntu-latest, 20|22|24|26),test (macos-latest, …)) andVercel Preview Comments. Zero contain "dco" — or "cla", "license", "sign" or "legal", for that matter. - Commit statuses (the older API, where DCO bots usually post): only
VercelandCodeRabbit.
The remaining uncertainty is narrower than I first thought, but it is not zero. Required status checks and rulesets configured in repository settings are not visible through either API without admin access, and a reviewer may simply be checking sign-off by eye. So the question rather than a claim: is sign-off enforced somewhere I cannot see, and is the requirement in CONTRIBUTING.md describing something that is checked, or describing intended practice?
If it is enforced by an app, it might help future contributors for CONTRIBUTING.md to name where. If it is not currently enforced, a DCO action would make the stated rule checkable — or the requirement could be softened to match practice.
How to reproduce
As of 2026-09-01:
gh pr list -R rohitg00/agentmemory --state merged --limit 20 \
--json number --jq '.[].number' |
while read N; do
c=$(gh api "repos/rohitg00/agentmemory/pulls/$N/commits" --jq '.[].sha' | wc -l)
s=$(gh api "repos/rohitg00/agentmemory/pulls/$N/commits" --jq '.[].commit.message' \
| grep -ci '^Signed-off-by:')
echo "PR $N: $c commits, $s signed"
done
ls -a .github/ && grep -rniE "dco|sign-off|signoff" .github/
# what actually ran on those PRs
gh pr list -R rohitg00/agentmemory --state merged --limit 20 --json number --jq '.[].number' |
while read N; do
sha=$(gh api "repos/rohitg00/agentmemory/pulls/$N" --jq .head.sha)
gh api "repos/rohitg00/agentmemory/commits/$sha/check-runs" --jq '.check_runs[].name'
gh api "repos/rohitg00/agentmemory/commits/$sha/status" --jq '.statuses[].context'
done | sort -uNote on method: git log -100 --grep="Signed-off-by" is not a reliable way to measure this — -100 is applied after the filter rather than to the window being examined, so it silently reports on a different set of commits than intended. Counting on the PR commits, as above, measures the rule where it is stated to apply.
Activity