smoke

name	smoke
description	Run a fresh Sprints smoke test for playground by syncing main, reinstalling the Sprints plugin, selecting the workflow root, applying the active workflow contract, starting managed daemons, creating or selecting a smoke issue, and observing until completion or operator attention.

Smoke

Use this skill when the user asks to start or observe a Sprints smoke test for attmous/playground.

Prefer the managed daemon path. Use foreground one-shot ticks only when the managed systemd user services are unavailable or the user explicitly asks for manual observation.

Inputs To Resolve

• Playground checkout: /home/radxa/WS/playground
• Sprints checkout: /home/radxa/WS/sprints
• Workflow name: default change-delivery
• Workflow root: prefer .hermes/sprints/workflow-root

If .hermes/sprints/workflow-root is missing, bootstrap or scaffold the workflow only after confirming the intended repo slug and workflow name.

Procedure

1. Sync both repositories to origin/main.

git -C /home/radxa/WS/sprints status --short --branch
git -C /home/radxa/WS/sprints pull --ff-only origin main
git -C /home/radxa/WS/playground status --short --branch
git -C /home/radxa/WS/playground switch main
git -C /home/radxa/WS/playground pull --ff-only origin main

If either checkout has local modifications, stop and report the dirty paths unless the user already asked to discard or preserve them.

2. Reinstall the Sprints plugin from the synced Sprints checkout.

python3 /home/radxa/WS/sprints/scripts/install.py

3. Pick and validate the workflow.

WORKFLOW_ROOT="$(cat /home/radxa/WS/playground/.hermes/sprints/workflow-root)"
python3 /home/radxa/WS/sprints/sprints_cli.py validate --workflow-root "$WORKFLOW_ROOT" --format text
python3 /home/radxa/WS/sprints/sprints_cli.py apply-contract --workflow-root "$WORKFLOW_ROOT" --source-ref origin/main --format text
python3 /home/radxa/WS/sprints/sprints_cli.py status --workflow-root "$WORKFLOW_ROOT" --format json

If apply-contract refuses because active lanes exist, use the clean skill first or ask the user whether to force the contract. Do not use --force silently.

4. Start or restart the managed services.

python3 /home/radxa/WS/sprints/sprints_cli.py codex-app-server up --workflow-root "$WORKFLOW_ROOT"
python3 /home/radxa/WS/sprints/sprints_cli.py daemon up --workflow-root "$WORKFLOW_ROOT"
python3 /home/radxa/WS/sprints/sprints_cli.py codex-app-server status --workflow-root "$WORKFLOW_ROOT" --format json
python3 /home/radxa/WS/sprints/sprints_cli.py daemon status --workflow-root "$WORKFLOW_ROOT" --json

If the managed daemon path is unavailable, use foreground one-shot ticks:

python3 /home/radxa/WS/sprints/sprints_cli.py daemon run --workflow-root "$WORKFLOW_ROOT" --once --json

Repeat one-shot ticks until the lane completes, enters operator_attention, or there is no runnable work.

5. Create or select the smoke issue.

If the user supplied an issue, ensure it has the workflow-required label active.

For a new issue, create a small scoped change and add active plus sprints-smoke.

gh issue create --repo attmous/playground --title "<smoke title>" --label active --label sprints-smoke --body '<issue body>'

Avoid Markdown backticks in double-quoted shell strings. Prefer single quotes or a temporary body file when the body contains code.

6. Observe.

Poll the workflow and tracker state:

python3 /home/radxa/WS/sprints/sprints_cli.py status --workflow-root "$WORKFLOW_ROOT" --format json
python3 /home/radxa/WS/sprints/sprints_cli.py events --workflow-root "$WORKFLOW_ROOT" --limit 20 --format text
gh issue list --repo attmous/playground --state open --label active --limit 20 --json number,title,url,labels
gh pr list --repo attmous/playground --state open --limit 20 --json number,title,url,headRefName,statusCheckRollup

When a PR appears, inspect it:

gh pr view <number> --repo attmous/playground --json number,title,state,url,headRefName,headRefOid,mergeable,isDraft,statusCheckRollup

7. Report outcome.

Include:

• Sprints commit used and whether plugin reinstall succeeded
• playground commit used
• workflow root and workflow name
• managed daemon/app-server status, or one-shot fallback used
• smoke issue URL
• PR URL, branch, commit, CI/check status
• final lane status: complete, running, retry, or operator_attention
• any anomalies observed, especially stale status fields, daemon lease behavior, or completed runs with misleading error fields
• lesson file written by the lessons skill under docs/lessons_learned/

Before sending the final smoke report, run the repo-local lessons skill and document the run's issue, resolution, lessons learned, and follow-up items. If the smoke had no anomalies, still add a short note saying that no new lesson was found and point to the final status checks.

When Starting Over

Run the clean skill first when the user asks to clean before a new smoke. That skill stops services, closes open PRs, and removes active labels without deleting workflow contracts or runtime state unless explicitly asked.