fix(mention): paginate gh api calls in verify job#228
Merged
Conversation
Without --paginate, gh api silently truncates at 30 results. On issues/PRs with many comments or reviews, this could miss bot engagement and skip a valid mention. Co-Authored-By: Claude <noreply@anthropic.com>
3 tasks
max-sixty
pushed a commit
that referenced
this pull request
Apr 10, 2026
#234) ## Summary Every `/tend-ci-runner:review` invocation has been **silently broken** since PR #217 merged at 2026-04-10 20:07 UTC. PRs #225–#233 all received **zero reviews from the bot** — the review runs report `success` but the session artifacts show no assistant turn at all. This PR removes the poison and restores reviews. ## Root cause PR #217 (`fix(ci-runner): replace if-not polling loop with || to avoid bash escaping`) added an explanatory comment inside the CI-polling code block: ```markdown # Use `||` rather than `if \!` — the Bash tool escapes `\!` as `\\!`, which # prevents bash from recognizing the pipeline-negation reserved word... ``` Claude Code's slash-command preprocessor scans the loaded file for `` \!` `` (a prefix that means "execute the next backticked string as a shell command") **regardless of markdown code fences**. It found `` \!` `` inside `` `if \!` ``, consumed text through the next backtick, and ran `` — the Bash tool escapes ` `` as bash. The em-dash is not a command, so bash reported: ``` Error: Shell command failed for pattern "\!` — the Bash tool escapes `": /bin/bash: line 1: —: command not found ``` When the preprocessor fails, the prompt body is not delivered. Every `/review <N>` session in this window produced the same 5-line trace — two internal `queue-operation` messages, the command caveat, the command invocation, the stderr — and then nothing. The model was never given instructions to follow. ## Evidence All tend-review sessions in the past hour's window ran this preprocessor and failed identically: | Run | PR | Branch | Session size | Review posted | |---|---|---|---|---| | 24261252532 | #224 | `review-runs-timeout-detection` | ~120 lines | ✅ APPROVED (ran **before** #217 merged at 20:07Z) | | 24262024783 | #225 | `tend-mention-dedup-review-events` | 5 lines | ❌ | | 24262028666 | #226 | `hourly/review-24261599977` | — | ❌ (checkout failure — infra race, separate issue) | | 24262305841 | #226 | `hourly/review-24261599977` | 5 lines | ❌ | | 24262406137 | #227 | `raise-job-timeouts` | 5 lines | ❌ | | 24262970448 | #228 | `mention-verify-paginate` | 5 lines | ❌ | | 24262987799 | #229 | `tend-claude-md-workflow-update-note` | 5 lines | ❌ | | 24263043865 | #230 | `raise-job-timeouts` | 5 lines | ❌ | | 24263117206 | #231 | `mention-review-inline-reply-prompt` | 5 lines | ❌ | | 24263177723 | #232 | `raise-job-timeouts` | 5 lines | ❌ | | 24263220447 | #233 | `tend-mention-dedup-review-events` | 5 lines | ❌ | Sample JSONL from run 24262987799 (PR #229): ```json {"type":"queue-operation"} {"type":"queue-operation"} {"type":"user","content":"<local-command-caveat>..."} {"type":"user","content":"<command-name>/review</command-name>\n<command-args>229</command-args>"} {"type":"user","content":"<local-command-stderr>Error: Shell command failed for pattern \"\!` — the Bash tool escapes `\": [stderr]\n/bin/bash: line 1: —: command not found</local-command-stderr>"} ``` No assistant turn. No Skill loads. No tool calls. No review. Confirmation via the target PRs: `gh pr view <N> --json reviews` returns `[]` for every one of PRs #225–#233. PR #224, reviewed before #217 landed, has the bot's APPROVED review. The same preprocessor error is also visible in the `tend-mention` run 24261917519 when it loaded `running-in-ci` via the Skill tool — there the error appeared as a tool_result and the model continued working, because skill-load failures are non-fatal. Only the top-level slash-command invocation is lethal. ## Gate assessment - **Evidence level**: Critical (wrong outcome — reviews 100% non-functional) - **Classification**: Structural (deterministic; every invocation fails the same way) - **Occurrences this run**: 9 (plus 1 non-fatal skill-load collision in tend-mention) - **Historical evidence**: N/A — this regression was introduced by #217 earlier today and has no prior tracking entries - **Change type**: Targeted fix (two-line comment rephrase in two files) - **Both gates**: pass ## Fix Rephrase the comments in `plugins/tend-ci-runner/skills/review/SKILL.md:294` and `plugins/tend-ci-runner/skills/running-in-ci/SKILL.md:142` to avoid any `` \!` `` sequence. The guidance content is unchanged: "Use `||` rather than `if`-based negation; the Bash tool escapes the exclamation mark to a literal backslash-exclamation." No `\!` now appears adjacent to a backtick anywhere in `plugins/`. `grep '\!\`' plugins/` after the fix returns zero matches. ## Test plan - [ ] A tend-review run on this PR completes with an assistant session longer than 5 lines and posts an actual review (self-approval blocked, but the bot must at least read the diff and comment) - [ ] `grep -r '\!\`' plugins/` returns no matches - [ ] `tend-mention` and `tend-notifications` skill loads of `running-in-ci` no longer show `Shell command failed for pattern` in tool_results 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: continuous-bot <269947486+continuous-bot@users.noreply.github.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
tend-agent
added a commit
that referenced
this pull request
Apr 11, 2026
Syncs the checked-in workflow file with the in-repo generator. Pulls in changes from #225 (pull_request_review_comment dedup), #233 (drop bot-author filter), #228 (--paginate on gh api), and #235 (move behavioral instructions from prompt to skill). Generated with the local `generator/` source tree rather than `uvx tend@latest init`, because PyPI 0.0.11 lags behind main for these changes — running `uvx tend@latest init` against the current tree would *add* `timeout-minutes: 60` back and drop the mention generator updates.
This was referenced Apr 11, 2026
tend-agent
pushed a commit
that referenced
this pull request
Apr 13, 2026
When `generator/pyproject.toml` exists (i.e. on the tend repo itself), Step 5 of the nightly skill now regenerates workflows via the in-tree generator instead of `uvx tend@latest`. Adopter repos without `generator/` fall through to the existing `uvx tend@latest` path unchanged. Motivation: PR #264 (from run 24330194038) ran `uvx tend@latest init` with PyPI's `tend` at 0.0.11 (released 2026-04-10). `main` has 10 generator commits landed since — #223 (drop hardcoded job timeouts), #225 (dedup review events), #227 (workflow_extra), #228 (paginate), #231 (inline-reply guidance), #233 (drop bot-author filter on issue_comment), #235 (prompt → skill), #254 (notifications freshness gate + bot-closed-PR drop), etc. Regenerating against the stale wheel reverted each of these. Maintainer flagged the PR as regressions; diagnosis confirmed the PyPI lag. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1 task
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Without
--paginate,gh apisilently truncates at 30 results. On issues/PRs with many comments or reviews, the verify job could miss bot engagement and skip a valid mention.Adds
--paginateto all fourgh apiGET calls in the verify step. The jq expressions work unchanged on paginated output.