local-e2e

name	local-e2e
description	Run a full local E2E test through the orchestrator-daemon pipeline. Starts infrastructure and services if needed, fires a webhook, tails logs until completion.

User Input

$ARGUMENTS

Required Environment Variables

These must be set in .env or the shell environment before running:

• LOCAL_E2E_TEST_REPO: owner/repo to test against (e.g. myorg/my-test-repo)
• LOCAL_E2E_TEST_ENTITY: issue or PR number to target

The skill MUST fail with a clear error message if either is unset.

Arguments

• $ARGUMENTS text becomes the trigger body (default: @chrisleekr-bot run docker compose to execute the test suite in this repo. After running, save any gotchas or environment notes you discovered using save_repo_memory.)
• If $ARGUMENTS contains --pr, set IS_PR=true and strip the flag from the trigger body
• If $ARGUMENTS contains --dry-run, set DRY_RUN=true and strip the flag from the trigger body
• Default mode is live execution (DRY_RUN=false)

Execution Steps

Step 1: Validate environment

1. Read LOCAL_E2E_TEST_REPO and LOCAL_E2E_TEST_ENTITY from the environment (check .env file if not in shell env).

2. If either is missing, stop immediately and tell the user:

   Missing required env vars for /local-e2e:
     LOCAL_E2E_TEST_REPO: set to owner/repo (e.g. myorg/my-test-repo)
     LOCAL_E2E_TEST_ENTITY: set to issue or PR number
   Add them to .env or export in your shell.
   

3. Split LOCAL_E2E_TEST_REPO into OWNER and REPO on the / delimiter.

Step 2: Start infrastructure if needed

1. Check if Docker containers for this project's Valkey and Postgres are running (docker ps).
2. If not running, execute bun run dev:deps and wait for containers to be healthy.

Step 3: Start services if needed

1. Check if the server is listening: curl -s http://localhost:3000/healthz
2. If not, start bun run dev in background. Wait for /healthz to return ok.
3. Check daemon logs or server logs for daemon registration.
4. If no daemon is connected, start bun run dev:daemon in background. Wait for Registered with orchestrator in daemon output.

Step 4: Fire the test

Run the test using the existing script:

DRY_RUN=${DRY_RUN:-false} \
OWNER="${OWNER}" \
REPO="${REPO}" \
ENTITY="${LOCAL_E2E_TEST_ENTITY}" \
IS_PR="${IS_PR:-false}" \
bash scripts/test-webhook.sh "${TRIGGER_BODY}"

Step 5: Monitor execution

1. Tail the daemon log output, watching for either:
   • Claude Agent SDK execution completed: success
   • error / failed / process exit: failure
2. Check every 15 seconds, with a maximum wait of 5 minutes.

Step 6: Report result

Print a summary:

• Status: success or failure
• Duration: from daemon logs (durationMs)
• Cost: from daemon logs (costUsd)
• Turns: from daemon logs (numTurns)
• Target: OWNER/REPO#ENTITY
• Mode: live or dry-run
• If live and successful, suggest checking the GitHub issue/PR for the posted comment.

Step 7: Verify memory persistence

1. Query repo_memory table: psql $DATABASE_URL -c "SELECT id, repo_owner, repo_name, category, content, pinned FROM repo_memory ORDER BY updated_at DESC LIMIT 10;"
2. Report whether new entries were created during this execution.
3. If empty, flag as memory not working: the daemon either didn't pass REPO_MEMORY env to the MCP server, the MCP server didn't write .daemon-actions.json, or the orchestrator didn't persist the actions.