GIGO · Lobster Doctor

🦞 GIGO · gigo-lobster-doctor: 环境体检模式：只检查 gateway、Python 依赖、题包链路与 PNG 证书能力，不跑正式试吃。 Triggers: 龙虾体检 / 检查龙虾环境 / lobster doctor / check lobster environment.

Security Scan

The skill's code, runtime instructions, and requirements are coherent with a local ‘environment doctor’ for the GIGO lobster benchmark; nothing requests unrelated credentials or unexplained installs, but it can perform runtime package bootstrapping and network checks which you should expect before running.

Purpose & Capability

Name/description (environment doctor) aligns with what is present: Python wrapper, doctor mode, checks for gateway, task bundle access, and PNG certificate capability. Declared binary requirement (python3/python/py) matches expected behavior. Files for report/cert generation, task fetching, and gateway interaction are present and appropriate for a suite that implements both 'doctor' and full tasting modes.

Instruction Scope

SKILL.md instructs the agent to run the wrapper (run_doctor.py) and to tail logs under the OpenClaw workspace; that is consistent with a runtime check. The SKILL.md also contains many authoritative runtime directives (e.g., 'never cd ... && python ...', 'do not inspect the whole repo'), which are intended to shape how the agent executes the wrapper but are effectively prompt-like controls. The instructions reference user-home workspace paths and log files (expected for this skill) and do not ask for unrelated secrets or system config files.

Install Mechanism

There is no external install spec in manifest.json, but the packaged code contains a runtime bootstrap (scripts/runtime_bootstrap.py) and a requirements.lock listing Pillow, qrcode, cryptography. The README and code indicate the skill can self-boot (pip/venv) to satisfy PNG certificate dependencies — this means network installs from PyPI may occur at runtime when ensure_runtime runs. That behavior is proportionate to the stated need (PNG generation) but is a runtime install action the user should expect.

Credentials

The skill does not declare required credentials or sensitive env vars. It reads optional vars (GIGO_*, GIGO_REQUIRE_PNG_CERT, GIGO_GATEWAY_MOCK) that map to its functionality. The code will contact a Gateway and cloud endpoints when not run with --offline; that is consistent with its task-bundle and version-check features and not disproportionate to the described purpose.

Persistence & Privilege

Skill is not force-included (always: false) and uses the normal model-invocation defaults. It does not request system-wide configuration nor claim to modify other skills. It writes output under the OpenClaw workspace/output paths, which is expected and declared.

Scan Findings in Context

[unicode-control-chars] unexpected: The SKILL.md contains directive-like phrasing and the scanner flagged unicode-control characters. The presence of control characters is unusual and could be an artifact of formatting or an attempt to shape agent behavior; however, the instructions themselves are consistent with runtime expectations for a wrapper-based skill. Consider inspecting the SKILL.md raw bytes before trusting unreviewed copies.

Assessment

This skill appears to be what it says: a local environment 'doctor' for the GIGO lobster benchmark. Before running it: 1) review scripts/runtime_bootstrap.py to confirm whether it will pip-install packages and from where (network fetches from PyPI are likely if Pillow/qrcode are missing); 2) run in --offline mode if you want to avoid network calls while checking local capabilities; 3) inspect SKILL.md and any non-PyPI endpoints in gateway_client/session_client if you have strict network/security constraints; and 4) check the flagged unicode-control characters in SKILL.md (open the file in a hex-capable editor) if you distrust prompt-like directives. If you do not want the skill to reach cloud endpoints, run run_doctor.py with --offline (the skill explicitly supports offline checks).

Mission

• 环境体检模式：只检查 gateway、Python 依赖、题包链路与 PNG 证书能力，不跑正式试吃。
• Environment doctor mode: checks the gateway, Python/runtime dependencies, task-bundle access, and PNG certificate readiness without running the full benchmark.

Trigger Phrases

• 中文：龙虾体检 / 检查龙虾环境 / 先体检龙虾 / 龙虾环境检查
• English: lobster doctor / check lobster environment / lobster environment check / doctor my lobster

Execution Rules

1. Use a direct Python command on this skill directory's wrapper file. Never use cd ... && python ...; OpenClaw preflight may reject it.
2. Prefer python3, then python, then py.
3. If the user asked in Chinese, append --lang zh. If the user asked in English, append --lang en.
4. Stream short progress updates while the benchmark is running.
5. Keep stdout/stderr visible and remind the user that the full log is written to gigo-run.log.
6. Do not run --help, inspect the whole repo, or switch to main.py once the wrapper command is clear. Start the wrapper directly.
7. If the wrapper starts a long-running process, do not kill it just because stdout is quiet for a while. A full tasting run often takes 15-25 minutes.
8. While a long run is in progress, monitor the process and tail the log file under ~/.openclaw/workspace/outputs/gigo-lobster-doctor/gigo-run.log instead of improvising a second execution path.
9. Only declare failure if the process exits non-zero, the log shows a traceback, or the user explicitly asks to cancel.
10. Stay attached until the wrapper exits. Do not end the conversation with “I will keep monitoring”; keep polling and only report completion once you have the final score/result files/ref_code (if any).
11. Prefer process poll plus exec tail -n 50 .../gigo-run.log while monitoring. Do not use a generic full-file read on gigo-run.log, because the log can be large and may break the chat output.

Default Behavior

• 中文：默认只做环境检查，不跑正式 benchmark，也不会上传。
• English: By default it only runs the environment checks. No full benchmark and no upload.

Recommended Command Shape

python3 /absolute/path/to/run_doctor.py --lang zh

If the user explicitly asks for overrides, append the matching CLI flags:

• --lobster-name "..." and --lobster-tags "tag1,tag2" for a custom lobster persona
• --output-dir /custom/path for a custom output directory
• --require-png-cert when the user refuses the SVG fallback
• --skip-upload or --register-only only when the user explicitly asks to change the default upload behavior

Persona Defaults

• Prefer SOUL.md first
• Then read GIGO_LOBSTER_NAME and GIGO_LOBSTER_TAGS
• Finally accept explicit CLI overrides

Do not stop for interactive questions unless the user explicitly asks for an interactive run.