E2E plugin & Add local plugin marketplace

[alishakawaguchi]

Summary

• Creates a local marketplace at .claude/plugins/ wrapping the e2e and agent-integration plugins, registered via extraKnownMarketplaces in project settings
• Adds e2e plugin with subcommands and skills for E2E test triage workflows
• Updates agent-integration skills

This makes plugin subcommands discoverable by Claude Code for all team members.

New E2E commands

Command	Description
/e2e	Orchestrator — runs triage-ci then implement sequentially
/e2e:triage-ci	Run failing E2E tests locally, classify flaky vs real-bug, present findings report
/e2e:debug	Deep-dive artifact analysis for diagnosing E2E test failures
/e2e:implement	Apply fixes from triage/debug findings, verify with scoped E2E tests

Agent integration commands (existing, now discoverable as subcommands)

Command	Description
/agent-integration	Run all three phases sequentially
/agent-integration:research	Assess a new agent's hook/lifecycle compatibility with the CLI
/agent-integration:write-tests	Create E2E agent runner (no unit tests)
/agent-integration:implement	E2E-first TDD — unit tests written last

Test plan

• Start a new Claude Code session in this project
• Verify /e2e: tab-completion shows debug, triage-ci, implement
• Verify /agent-integration: tab-completion shows research, write-tests, implement
• Run /e2e:debug and verify it loads the procedure

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Documentation and Claude Code configuration changes only; no production/runtime code paths are affected.

Overview
Adds a local Claude Code marketplace (.claude/plugins) and enables it via .claude/settings.json, making the e2e and agent-integration plugins available to the project.

Introduces an e2e plugin with discoverable commands (/e2e:triage-ci, /e2e:debug, /e2e:implement) backed by new E2E triage/debug/implement skill docs, and updates agent-integration skill documentation to reference the new /e2e:debug command instead of the old /debug-e2e.

^{Written by Cursor Bugbot for commit 994602b. Configure here.}

[Copilot]

Pull request overview

Adds a local Claude Code plugin marketplace under .claude/plugins/ and wires up new /e2e:* commands plus updated e2e + agent-integration skills to make subcommands discoverable for the whole team.

Changes:

• Register a local marketplace via .claude/settings.json and enable e2e + agent-integration plugins from it
• Add e2e plugin command wrappers (/e2e:debug, /e2e:triage-ci, /e2e:implement) pointing at new/updated skill procedures
• Update agent-integration docs to reference /e2e:debug instead of /debug-e2e

Reviewed changes

Copilot reviewed 14 out of 14 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
.claude/settings.json	Registers local plugin marketplace and enables the new local plugins
.claude/plugins/.claude-plugin/marketplace.json	Defines the local marketplace and the plugins it exposes
.claude/plugins/e2e/.claude-plugin/plugin.json	Declares the e2e plugin metadata for marketplace loading
.claude/plugins/e2e/commands/debug.md	Adds /e2e:debug command wrapper pointing to the debug skill
.claude/plugins/e2e/commands/triage-ci.md	Adds /e2e:triage-ci command wrapper pointing to the triage skill
.claude/plugins/e2e/commands/implement.md	Adds /e2e:implement command wrapper pointing to the implement skill
.claude/plugins/e2e/README.md	Documents the newly added e2e plugin commands
.claude/skills/e2e/SKILL.md	Adds an orchestrator skill that runs triage then implement
.claude/skills/e2e/debug.md	Introduces the E2E artifact debugging workflow
.claude/skills/e2e/triage-ci.md	Introduces the CI triage workflow with rerun-based classification
.claude/skills/e2e/implement.md	Introduces the “implement fixes” workflow with mandatory real E2E verification
.claude/skills/agent-integration/SKILL.md	Updates references to use /e2e:debug
.claude/skills/agent-integration/implementer.md	Updates multiple workflow instructions to use /e2e:debug
.claude/skills/agent-integration/test-writer.md	Updates debugging guidance to use /e2e:debug

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

Bugbot Autofix prepared fixes for both issues found in the latest run.

• ✅ Fixed: Triage procedure references non-existent download script
  • Added scripts/download-e2e-artifacts.sh to resolve latest/run-id/run-url references and output the artifact directory path as expected.
• ✅ Fixed: New debug skill duplicates existing debug-e2e skill
  • Replaced .claude/skills/e2e/debug.md with a pointer to .claude/skills/debug-e2e/SKILL.md so there is a single authoritative workflow.

Or push these changes by commenting:

@cursor push c57d165f1b

Preview (c57d165f1b)

diff --git a/.claude/skills/e2e/debug.md b/.claude/skills/e2e/debug.md
--- a/.claude/skills/e2e/debug.md
+++ b/.claude/skills/e2e/debug.md
@@ -1,88 +1,3 @@
-# Debug Entire CLI via E2E Artifacts
+# E2E Debug
 
-Diagnose Entire CLI bugs using captured artifacts from the E2E test suite. Artifacts are written to `e2e/artifacts/` locally or downloaded from CI via GitHub Actions.
-
-## Inputs
-
-The user provides either:
-- **A test run directory:** `e2e/artifacts/{timestamp}/` -- triage all failures
-- **A specific test directory:** `e2e/artifacts/{timestamp}/{TestName}-{agent}/` -- debug one test
-
-## Artifact Layout
-
-```
-e2e/artifacts/{timestamp}/
-├── report.nocolor.txt          # Pass/fail/skip summary with error lines
-├── test-events.json            # Raw Go test events (NDJSON)
-├── entire-version.txt          # CLI version under test
-└── {TestName}-{agent}/
-    ├── PASS or FAIL            # Status marker
-    ├── console.log             # Full operation transcript
-    ├── git-log.txt             # git log --decorate --graph --all
-    ├── git-tree.txt            # ls-tree HEAD + checkpoint branch
-    ├── entire-logs/entire.log  # CLI structured JSON logs
-    ├── checkpoint-metadata/    # Checkpoint + session metadata
-    └── repo -> /tmp/...        # Symlink to preserved repo (E2E_KEEP_REPOS=1 only)
-```
-
-## Preserved Repo
-
-When the test run was executed with `E2E_KEEP_REPOS=1`, each test's artifact directory contains a `repo` symlink pointing to the preserved temporary git repository. This is the actual repo the test operated on -- you can inspect it directly.
-
-**Navigate via the symlink** (e.g., `{artifact-dir}/repo/`) rather than resolving the `/tmp/...` path. The symlink lives inside the artifact directory so permissions and paths stay consistent.
-
-The preserved repo contains:
-- Full git history with all branches (main, `entire/checkpoints/v1`)
-- The `.entire/` directory with CLI state, config, and raw logs
-- The `.claude/` directory (if Claude Code was the agent)
-- All files the agent created or modified, in their final state
-
-This is the most powerful debugging tool -- you can run `git log`, `git diff`, `git show`, inspect `.entire/` internals, and see exactly what the CLI left behind.
-
-## Debugging Workflow
-
-### 1. Triage (if given a run directory)
-
-Read `report.nocolor.txt` to identify failures and their error messages. Each entry shows the test name, agent, duration, and failure output with file:line references.
-
-### 2. Read console.log (most important)
-
-Full transcript of every operation:
-- `> claude -p "..." ...` -- agent prompts with stdout/stderr
-- `> git add/commit/...` -- git commands
-- `> send: ...` -- interactive session inputs
-
-This tells you what happened chronologically.
-
-### 3. Read test source code
-
-Use the file:line from the report to find the test in `e2e/tests/`. Understand what the test expected to happen vs what console.log shows actually happened.
-
-### 4. Diagnose the CLI behavior
-
-Cross-reference console.log (what happened) with the test (what should have happened). Focus on CLI-level issues:
-
-| Symptom | CLI Investigation |
-|---------|-------------------|
-| Checkpoint not created / timeout | Check `entire-logs/entire.log` for hook invocations, phase transitions, errors |
-| Wrong checkpoint content | Check `git-tree.txt` for checkpoint branch files, `checkpoint-metadata/` for session info |
-| Hooks didn't fire | Check `entire.log` for missing hook entries (session-start, user-prompt-submit, stop, post-commit) |
-| Stash/unstash problems | Check `entire.log` for stash-related log lines, `git-log.txt` for commit ordering |
-| Attribution issues | Check `checkpoint-metadata/` for `files_touched`, session metadata for attribution data |
-| Strategy mismatch | Check `entire.log` for `strategy` field, verify auto-commit vs manual-commit behavior |
-
-### 5. Deep dive files
-
-- **entire-logs/entire.log**: Structured JSON logs -- hook lifecycle, session phases (`active` -> `idle` -> `ended`), warnings, errors. Key fields: `component`, `hook`, `strategy`, `session_id`.
-- **git-log.txt**: Commit graph showing main branch, `entire/checkpoints/v1`, checkpoint initialization.
-- **git-tree.txt**: Files at HEAD vs checkpoint branch (separated by `--- entire/checkpoints/v1 ---`).
-- **checkpoint-metadata/**: `metadata.json` has `checkpoint_id`, `strategy`, `files_touched`, `token_usage`, and `sessions` array. Session subdirs have per-session details.
-
-### 6. Report findings
-
-Identify whether the issue is in:
-- **CLI hooks** (prepare-commit-msg, commit-msg, post-commit)
-- **Session management** (phase transitions, session tracking)
-- **Checkpoint creation** (branch management, metadata writing)
-- **Attribution** (file tracking, prompt correlation)
-- **Strategy logic** (auto-commit vs manual-commit behavior)
+Use `.claude/skills/debug-e2e/SKILL.md` as the single source of truth for the debugging workflow.

diff --git a/scripts/download-e2e-artifacts.sh b/scripts/download-e2e-artifacts.sh
new file mode 100644
--- /dev/null
+++ b/scripts/download-e2e-artifacts.sh
@@ -1,0 +1,45 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+usage() {
+	echo "Usage: $0 <latest | RUN_ID | RUN_URL>" >&2
+	exit 1
+}
+
+if [[ $# -ne 1 ]]; then
+	usage
+fi
+
+if ! command -v gh >/dev/null 2>&1; then
+	echo "Error: gh CLI is required but not installed." >&2
+	exit 1
+fi
+
+ref="$1"
+run_id=""
+
+case "$ref" in
+latest)
+		run_id="$(gh run list --workflow e2e.yml --branch main --status failure --limit 1 --json databaseId --jq '.[0].databaseId')"
+		;;
+	*)
+		if [[ "$ref" =~ ^[0-9]+$ ]]; then
+			run_id="$ref"
+		elif [[ "$ref" =~ /runs/([0-9]+) ]]; then
+			run_id="${BASH_REMATCH[1]}"
+		else
+			usage
+		fi
+		;;
+esac
+
+if [[ -z "$run_id" || "$run_id" == "null" ]]; then
+	echo "Error: Could not resolve a CI run ID from '$ref'." >&2
+	exit 1
+fi
+
+artifact_dir="$(mktemp -d "${TMPDIR:-/tmp}/entire-e2e-artifacts-${run_id}-XXXXXX")"
+
+gh run download "$run_id" --dir "$artifact_dir" >/dev/null
+
+printf '%s\n' "$artifact_dir"

_{This Bugbot Autofix run was free. To enable autofix for future PRs, go to the Cursor dashboard.}

Comment @cursor review or bugbot run to trigger another review on this PR