Conversation
…script Automates E2E failure triage with three new components: - scripts/download-e2e-artifacts.sh: reusable script to download CI artifacts - .claude/skills/e2e-triage/SKILL.md: 7-step triage skill (classify flaky vs real bug, create PRs or issues) - .github/workflows/e2e-triage.yml: workflow_run trigger that auto-runs Claude Opus on E2E failure Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 1aa72dcd8a2b
Post "Claude is triaging..." when triage starts and a structured summary with PR/issue links when it completes. The skill now writes triage-summary.json which the workflow parses with jq for the Slack message. Falls back to a warning if no summary is produced. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 8e5dcc6ef8ab
…ifications - Build Slack payload via jq (payload-file-path) instead of interpolating raw text into inline JSON, which broke on quotes/newlines in summaries - Add secrets.E2E_SLACK_WEBHOOK_URL guard to "Build Slack summary" and "Notify Slack - triage complete" steps (matching the "started" step) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 7c1914052967
Rewrite SKILL.md with dual-mode support (auto-detected via WORKFLOW_RUN_ID env var): local mode runs tests with mise and re-runs failures up to 3 times, CI mode triggers e2e-isolated.yml workflows for re-run verification. Classification now uses re-run results as the primary signal (all fail = real-bug, mixed results = flaky). Workflow changes: actions permission upgraded to write for gh workflow run, timeout increased to 60m for re-run polling, Claude prompt updated with CI mode hint and re-run instructions. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 9f75c3effd9b
Local mode now presents findings interactively and applies fixes directly in the working tree instead of creating branches/PRs/issues: - Step 4a: findings report, proposed fixes, user approval gate, in-place fixes - Step 4b: unchanged CI behavior (batched PR for flaky, issues for real bugs) - Step 5: local mode gets simpler summary table, no triage-summary.json Entire-Checkpoint: 4e1d9cf59d52
Consistent test failures can be test infrastructure bugs (e2e/ code), not product bugs (cmd/entire/cli/). Update classification signals, fix lists, and action sections to distinguish the two. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 28c90fcc7266
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: ba6877944a6c
Replace duplicated artifact-reading steps in e2e-triage Step 1 with a reference to debug-e2e's Debugging Workflow (steps 2-5), keeping the collect list so classification inputs remain clear. Add Related Skills section to README. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: eb14496bde1e
Entire-Checkpoint: bb778fbab533
…d GitHub issues - Remove workflow_run trigger from e2e-triage.yml (now dispatch-only) - Remove issues permission and gh issue commands from CI mode - Replace real-bug GitHub issues with structured CI log reports - Add triage link to Slack failure notification in e2e.yml - Update skill docs and README to reflect new behavior Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 4725d29fd8ff
Remove CI mode (branch creation, PRs, triage-summary.json, CI re-runs via gh workflow run) from the e2e-triage skill while preserving local debugging of CI failures (downloading artifacts, analyzing them, running tests locally). Also removes the e2e-triage.yml workflow and the triage link from the E2E Slack failure notification. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: c8ddf0fdb9df
Teach Step L1 to accept CI run references (latest, run ID, run URL) and use scripts/download-e2e-artifacts.sh to fetch artifacts, skipping local re-runs and jumping straight to shared analysis. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: fb0c122d807e
Skip re-downloading when the artifact directory already exists and is non-empty, printing a log message instead. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 5f75d4661043
…ge skill Pre-create ~/.config/cursor/ in Bootstrap() so the cursor CLI doesn't crash with ENOENT when writing cli-config.json after accepting workspace trust on CI. Follows the same pattern used by Claude, Gemini, and Droid agents. Update e2e-triage skill to require running real E2E tests after applying fixes, scoped by change type: agent-specific → that agent's full suite, shared infra → all affected agents, prompt-only → just the affected test. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: fa1e4e3fb457
Split the monolithic e2e-triage skill into three focused commands (triage-ci, debug, implement) following the agent-integration plugin pattern. Triage-ci is report-only, implement is action-only, and the /e2e orchestrator runs both sequentially. - Create .claude/plugins/e2e/ with plugin.json and command wrappers - Create .claude/skills/e2e/ with orchestrator SKILL.md and 3 procedures - Delete old .claude/skills/e2e-triage/ and .claude/skills/debug-e2e/ - Update all /debug-e2e references to /e2e:debug in agent-integration Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 2b582392c6b6
PR SummaryMedium Risk Overview Fixes a deferred-condensation edge case in manual commit hooks: Adds focused tests covering checkpoint ID recording on condensation failure and deferred checkpoint creation (including the “still empty transcript” no-crash path). Written by Cursor Bugbot for commit 6bbb43e. Configure here. |
Contributor
There was a problem hiding this comment.
Pull request overview
Adds tooling and documentation to improve debugging/triage of flaky E2E runs (especially Cursor), including a helper script for downloading CI artifacts and new .claude skill/plugin docs for a triage→fix workflow.
Changes:
- Add
scripts/download-e2e-artifacts.shto fetch and normalize GitHub Actions E2E artifacts locally. - Update Cursor E2E agent bootstrap to pre-create the Cursor config directory to avoid runtime ENOENT failures.
- Add/update E2E triage/debug/implement skill + plugin documentation and refresh E2E README guidance.
Reviewed changes
Copilot reviewed 15 out of 16 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
| scripts/download-e2e-artifacts.sh | New helper script to locate a run (latest/ID/URL), download artifacts, flatten wrapper dirs, and write .run-info.json. |
| e2e/agents/cursor_cli.go | Create Cursor config directory during bootstrap to reduce flaky failures. |
| e2e/README.md | Document local artifact download + new triage workflow references. |
| cmd/entire/cli/strategy/common_test.go | Minor formatting/alignment in tests. |
| .claude/skills/e2e/triage-ci.md | New CI triage procedure doc (download artifacts or rerun locally, classify flaky vs real-bug). |
| .claude/skills/e2e/implement.md | New procedure doc for applying fixes and verifying via scoped E2E runs. |
| .claude/skills/e2e/debug.md | Update/normalize debug procedure doc formatting. |
| .claude/skills/e2e/SKILL.md | Add orchestrator skill definition for the E2E triage→implement pipeline. |
| .claude/skills/agent-integration/*.md | Update references to use /e2e:debug instead of the old command name. |
| .claude/plugins/e2e/** | Add local plugin command wrappers + plugin metadata and README. |
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
Cursor's atomic config write (cli-config.json.tmp → cli-config.json)
races when parallel tests trigger "Workspace Trust Required"
simultaneously. Pre-seeding the file with {} in Bootstrap() avoids
the temp-file rename path entirely.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Entire-Checkpoint: 47c9b5f45145
The previous pre-seed fix wasn't sufficient because cursor's write-after-trust-acceptance still uses atomic temp-file rename on the shared ~/.config/cursor/cli-config.json. With 43 parallel tests, multiple cursor processes race on the same .tmp file. Fix: give each tmux session its own XDG_CONFIG_HOME + HOME pointing to an isolated temp directory with a pre-seeded cli-config.json, following the same pattern Claude Code uses with CLAUDE_CONFIG_DIR. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: df802d707cb9
- Add per-session XDG_CONFIG_HOME with pre-seeded cli-config.json to prevent ENOENT race on parallel tests (keep real HOME for auth) - Wait for "Add a follow-up" instead of PromptPattern() to avoid premature WaitFor settling during Thinking phase - Clean up temp config dir on session close Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: 121daaca3e20
When Cursor commits mid-turn before flushing its transcript, condensation fails silently and the checkpoint ID is never recorded. This means the commit trailer points to nonexistent checkpoint metadata. Fix by separating intent from execution: record checkpoint IDs on condensation *attempt* (not just success), and fall back to deferred CondenseSession at stop time when UpdateCommitted returns ErrCheckpointNotFound. No behavior change for agents that flush transcripts before committing (Claude Code, Gemini, etc.). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Entire-Checkpoint: fb5a3e2fb741
Creates a local marketplace at .claude/plugins/ wrapping both the e2e and agent-integration plugins, and registers it via extraKnownMarketplaces in project settings so plugin subcommands (/e2e:debug, /agent-integration:research, etc.) are discoverable by Claude Code. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Entire-Checkpoint: 13d3f0b6de12
- Standardize on `entire-logs/entire.log` in debug.md diagnostic table - Update triage-ci command description to mention CI artifact support Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Entire-Checkpoint: d9972bc12f95
alishakawaguchi
changed the title
Contributor
Author
|
@cursor review |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
alishakawaguchi
changed the title
khaong
approved these changes
Mar 17, 2026
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
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.
Summary
Two targeted fixes discovered during E2E triage:
~/.config/cursor/cli-config.jsonfile, causingENOENTerrors. Each session now gets an isolatedXDG_CONFIG_HOMEwith a pre-seededcli-config.json. Also switchesWaitForfrom the/ commandsprompt pattern (always visible in the status bar, even during "Thinking") to theAdd a follow-uptext that only appears after the agent finishes.🤖 Generated with Claude Code