| name | agentwheel-smoke |
| description | Run a read-only smoke diagnostic for the current AI harness and its Agentwheel-managed installation state. Use when asked to smoke test an agent, verify fleet deployment, inspect whether a harness is installed in the expected Agentwheel path, or summarize local Agentwheel status without mutating files or external systems. |
Agentwheel Smoke
Use this skill to answer: "What harness am I running in, where was this skill installed, and what does Agentwheel think about this target?"
The default workflow is read-only. Do not run agentwheel install, agentwheel update, agentwheel uninstall, agentwheel remember, agentwheel eject, or any file-writing fix while using this skill unless the user separately approves that mutation scope.
Quick Start
From the user's current workspace, run the bundled smoke script:
node path/to/agentwheel-smoke/scripts/smoke.mjs
Use JSON output only when another tool or script needs to consume the result:
node path/to/agentwheel-smoke/scripts/smoke.mjs --json
When the user names an Agentwheel target, pass only read-only status target flags:
node path/to/agentwheel-smoke/scripts/smoke.mjs --agent lab-codex
node path/to/agentwheel-smoke/scripts/smoke.mjs --profile daily
node path/to/agentwheel-smoke/scripts/smoke.mjs --all
Report The Result
Summarize the script output in four parts:
- Overall:
PASS, WARN, or FAIL.
- Harness: detected adapter, installation type, runtime root, and skill path.
- Agentwheel: CLI availability, version, config locations, packages, agents, profiles, and status result.
- Recommended next command: prefer
agentwheel status, agentwheel plan, or agentwheel install --dry-run.
Treat these as important signals:
PASS: skill path maps to a documented Agentwheel harness target and agentwheel status reports no pending work.WARN: the skill is running from package source, status is missing, config is absent, manifest is missing, or pending work is reported.FAIL: Agentwheel CLI is unavailable, the skill path maps to a known wrong target, agentwheel status exits unsuccessfully, or drift/conflict appears.
Compatibility Checks
Use these documented skill targets when explaining path compatibility:
- Codex local:
.agents/skills/<name>/SKILL.md; Codex user: ~/.agents/skills/<name>/SKILL.md.
- Claude local:
.claude/skills/<name>/SKILL.md; Claude user: ~/.claude/skills/<name>/SKILL.md.
- Copilot local:
.github/skills/<name>/SKILL.md; Copilot user: ~/.copilot/skills/<name>/SKILL.md.
- OpenClaw local:
skills/<name>/SKILL.md; OpenClaw user: ~/.openclaw/skills/<name>/SKILL.md.
- Hermes user:
~/.hermes/skills/<name>/SKILL.md; local Hermes skills are unsupported by default.
Flag .codex/skills/<name>/SKILL.md as incompatible for Agentwheel-managed Codex skills.
Safety
This skill is diagnostic. It may inspect files and run read-only Agentwheel commands, but it must not mutate runtime outputs, package source, Google Workspace data, Git state, registry state, or deployment state.
