chore: update tend workflows (0.2.5 → 0.2.6) - #627
Conversation
Deploying mouseterm with
|
| Latest commit: |
0a13f35
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://864f9b15.mouseterm.pages.dev |
| Branch Preview URL: | https://tend-update-workflows.mouseterm.pages.dev |
dormouse-bot
left a comment
There was a problem hiding this comment.
Feedback on work in progress — not a merge verdict. Mark ready when you want the full review.
One thing to land with the bump: docs/specs/security-ci.rationale.md still reads "The checked-in workflows use 0.2.5 as inspected in September 2026" under the 0.1.18 gap heading. After this PR the workflows pin 0.2.6, so the note contradicts the files in the same month it claims to describe. Bumping it alongside the regen is the established pattern here — commit 1d2c139 ("docs(security-ci): fork PR trees now land in the harness clone") did exactly that for the 0.2.5 bump. The 0.1.19 security floor in the same sentence is unaffected. The line is outside this PR's diff, so it can't be applied as an inline suggestion.
The checked-in tend workflows now pin 0.2.6, so the rationale's inspected-version note contradicted the files it describes. The 0.1.19 security floor is unchanged.
|
Applied in 0a13f35: the A repo-wide grep turned up no other stale |
Regenerates the tend workflows from tend 0.2.6 (was 0.2.5). The only generated change beyond the version stamps and the
max-sixty/tend/claude@action pin is intend-mention.yaml's verify script.A review the bot leaves on its own PR no longer dispatches a second session. Until now, a
pull_request_reviewauthored by the bot on a PR authored by the bot dispatched a "participation" run whenever the review carried a body or fresh inline comments — so the review run posted findings and a mention run then booted to act on them. Upstream moved that work into the review session itself (max-sixty/tend#1203), so the dispatch is now refused outright. Naming@dormouse-botinside a review still summons a session: the mention checks run ahead of this refusal. Reviews from humans, and reviews on PRs the bot didn't author, are unaffected.Two fixes land in the harness rather than the generated YAML. The CI sandbox now gets a writable temp directory (max-sixty/tend#1199) — this nightly run hit exactly that, with
/tmpread-only and every scratch path rerouted by hand. And the bundled skills' description ofauthor_associationis corrected to say the field is nested rather than absent (max-sixty/tend#1178), which is what the maintainer-tier gate reads before acting on a directive.Full comparison: 0.2.5...0.2.6
.config/tend.yamlis unchanged — this repo pins no harness model, so there was nothing to re-point at a newer default.Alongside the regen,
docs/specs/security-ci.rationale.mdgets its inspected-version note moved from0.2.5to0.2.6so it matches the pin these workflows now carry; the0.1.19security floor in that sentence is unchanged.