Repository navigation
Conversation
The repo-level secret is deleted; no workflow on any branch referenced it.
Probed against live GitHub: a pull_request_target run is admitted by a main-only policy, and a tag run is rejected unless the policy carries a tag entry. So most of these jobs can name the environment as they stand, and worktrunk's release jobs can too once the policy admits tags.
The smoke ran 4 times ever, all during the harness's construction week in June; ~15 harness changes since shipped without anyone dispatching it. Its deterministic half lives in CI's test-proxy job, and a broken release surfaces within a day through tend's own scheduled workflows dogfooding it. claude-smoke was tend-manual's only consumer, so the reviewer-gated environment goes with it. The credential-environments sweep stays: it is written against any reviewer-gated environment a repo declares, not the name, and the security model keeps documenting that pattern as the answer for bot-steerable triggers.
tend-agent
left a comment
There was a problem hiding this comment.
The removal itself checks out: TEND_MANUAL_ENVIRONMENT had no reader anywhere in the tree (only its definition), .github/claude-smoke-prompt.md had no other consumer, and the credential-environments sweep really is keyed on holding a credential rather than on a name, so docs/security-model.md's rewrite matches check_credential_environments. Generator tests and pre-commit both pass on the merged tree.
I checked the other two repos' workflows against the rewritten TODO item, and the ref accounting holds: worktrunk's benchmark gist append and both create-issue-on-*-failure jobs are if: … github.event_name == 'schedule', prql's update-rust-toolchain is gated on nightly-upstream, which is github.event_name == 'schedule' && github.repository_owner == 'prql', and prql's backport is the pull_request_target case. The remaining TEND_BOT_TOKEN readers in prql are release.yaml's brew-dispatcher, publish-winget, and push-web-branch — all covered by the third paragraph. One conflict with tend's own check, inline.
check_environment pins the tend policy to exactly the admitted branches and fix_environment deletes anything else, so the tag entry lives on a separate release environment, the shape check_credential_environments credits.
…cord guards (#835) ## Problem `review` derives three separate signals from the bot's own review records, and all three counted any record the bot owns. GitHub creates the *same* record shape for something that is not a review: replying to a review thread via `POST /pulls/{n}/comments/{id}/replies` wraps the reply in a synthetic zero-body `COMMENTED` review anchored at whatever HEAD was when the reply posted. So after any `tend-mention` run answers review threads inline — its normal response path — containers sit on the current HEAD, and every guard that asks "did the bot review this commit?" answers yes: - **`LAST_REVIEW_SHA`** (step 1) advances to a commit nothing reviewed, so the next `tend-review` run computes its incremental against code it never read and skips everything before it under the trivial-change heuristic. This is #828. - **`ALREADY_POSTED`** (step 5, *Posting mechanics*) becomes a false positive and the run exits at `Already reviewed — skipping`. Same lost-review-signal symptom, relocated to the close-out — and it costs more, because the run has already read the diff and formed the verdict before throwing it away. Fixing the derivation doesn't cover it: that routes the run *into* a full review, and this guard then discards the result. - **`ORPHAN_ID`** (*Recovering from inline comment 422 errors*) resolves to the container, so the case-(a) recovery would `PUT` the review body over an unrelated inline reply. Nothing goes red in any of the three; the run reports a well-reasoned decision to stay silent. On a repo where the bot both authors and reviews its own PRs, that removes the PR's only review signal. It is deterministic — no decision point — so it recurs on every reply-then-push flow. ## Solution Count a review record only when it carries a body, owns at least one *top-level* (non-reply) inline comment, or is `APPROVED`. Reply containers have none of the three; an empty-body review carrying real inline findings still anchors, and so does an empty-body approval. `LAST_REVIEW_SHA` and `ALREADY_POSTED` use the full predicate (the latter re-derives `SUBSTANTIVE`, since shell state doesn't carry between tool calls); `ORPHAN_ID` needs only the body test, because a case-(a) orphan always persisted a body and a container never has one. The review and comment calls are `--paginate`d — container churn is exactly what pushes a PR past one page of reviews — and each reduction runs over the merged array via `jq -rs 'add | …'`, not inside `--jq`: `gh` applies `--jq` per page, so a `| last` there yields one record *per page* rather than one overall. Where the two shapes are genuinely indistinguishable (a UI review whose only content is thread replies), the filter skips it — that moves `LAST_REVIEW_SHA` *earlier*, widening the incremental. The failure direction is more review, not less. ## Testing No test file covers skill prompt text, so the reproduction is a live API comparison rather than a committed test. **Step 1, against `max-sixty/cargo-affected` PR #54** — the case in #828: | review | `commit_id` | body | comments | |---|---|---|---| | 4851810736 | `1ffcde2` | 2441 chars | 2 top-level | | 4851851883 | `551e059` | 0 | 1 reply (`in_reply_to_id` 3710403279) | | 4851852052 | `551e059` | 0 | 1 reply (`in_reply_to_id` 3710403291) | The old recipe returns `551e059` — a commit with no review, and the one whose refactor got skipped. The new recipe returns `1ffcde2`, the real review. **Steps 5 and the 422 recovery, against this PR's own records.** `tend-mention` answered both threads from the first review after the `8699228` push, so the containers landed on the then-current HEAD: | record | `commit_id` | body | inline comments | |---|---|---|---| | 4860021420 | `b75b165` | 2286 | 2 top-level | | 4860032792 | `8699228` | 0 | 1 reply (`in_reply_to_id` 3717103568) | | 4860032847 | `8699228` | 0 | 1 reply (`in_reply_to_id` 3717103574) | | 4860059883 | `8699228` | 2812 | 0 | Evaluated against `8699228` as the reviewing run saw it — that is, before its own record 4860059883 existed — the old `ALREADY_POSTED` returns `2026-08-05T00:32:19Z` and the old `ORPHAN_ID` returns `4860032847`. Neither is a review; following the skill literally, that review would not have been posted at all. The new guards discard both containers: at `8699228` they return `2026-08-05T00:39:33Z` and `4860059883`, the one real review at that SHA, and at `7a11332` and `cdbf065` both return empty, where no review exists yet. The zero-body `APPROVED` records on #812 (`02f8dce`) and #820 (`3a5621c`, alongside two containers) both still anchor, which is what the `APPROVED` clause is for. Other consumers of the same records were checked and left alone: `weekly` filters on `state == "APPROVED"` (reply containers are always `COMMENTED`), and `notifications` counts recent bot reviews for dedup, where a reply container legitimately means the bot responded. --- Closes #828 — automated triage --------- Co-authored-by: tend-agent <270458913+tend-agent@users.noreply.github.com>
Automated nightly regeneration of tend's workflow files. **tend version:** 0.1.13 → 0.1.14 ## Notable changes - **Jobs now name the `tend` environment with `deployment: false`** (max-sixty/tend#852), so GitHub stops filing a deployment record per run and posting it on the pull request. This is what the current `tend check` flags as `environment-deployments` (#3729) — regenerating clears it, and max-sixty/tend#853 makes `check` refuse the old shape going forward. - **`id-token: write` dropped from every tend job** — a side effect of removing the claude-smoke workflow and its `tend-manual` environment (max-sixty/tend#820). None of the remaining jobs use OIDC, so the permission was unused. - **`tend-mention` counts bot engagement outside `jq`** (max-sixty/tend#840). `gh api --paginate` applies `--jq` once per page, so `| length` emitted one count per page; past 100 comments the shell variable held `100\n7`, the numeric test errored, and the bot fell through to `should_run=false` — going quiet on exactly its most-engaged threads. - **Review and triage skill fixes** — the review-record guards now ignore synthetic reply containers (max-sixty/tend#835), `/code-review` is ported into a tend-owned skill (max-sixty/tend#819), and triage substitutes the real issue number into its PR-body templates instead of leaving a placeholder (max-sixty/tend#844). - **Outage reporting is more robust** — a stranded outage row now names the trigger it points at (max-sixty/tend#823), and marking a notification read tolerates a transient run-metadata fetch failure (max-sixty/tend#843). Full compare: max-sixty/tend@0.1.13...0.1.14 Co-authored-by: worktrunk-bot <254187624+worktrunk-bot@users.noreply.github.com>
Removes the
claude-smokeworkflow and, once this merges, thetend-manualenvironment that existed only to feed it secrets.The smoke was scaffolding for building the headless proxy harness: it ran 4 times ever, all June 17–19 while #704–#712 were under construction, and none of the ~15 harness changes since (sandbox PATH, uv isolation, sandbox levers, removing
claude-interactive, the proxy readiness fix) dispatched it. Its deterministic half is covered per-PR by CI'stest-proxyjob at the production-pinned mitmproxy, and a harness broken in a release surfaces within a day through tend's own scheduled workflows dogfooding it, with a fix-forward release costing about an hour.tend-manualwas the reviewer-gated companion environmentclaude-smokeneeded because its whole point — dispatching on feature branches — is what thetendref policy exists to refuse. With its only consumer gone, the environment goes too, which also drops one paste from the credential migration inTODO.md. Thecredential-environmentssweep is untouched: it verifies any reviewer-gated environment a repo declares, keyed on the credential rather than the name (its tests assert exactly that), anddocs/security-model.mdkeeps documenting the pattern as the answer for bot-steerable triggers.Also carried on this branch: two TODO.md corrections from the migration work — dropping the deleted cargo-affected
BOT_TOKENleftover, and replacing the wrong claim that amain-only policy can never admit the worktrunk/prql hand-maintained workflows' refs (probed live:pull_request_targetruns atrefs/heads/mainand is admitted; a tag push is admitted once the policy carries a tag entry).