AI健身教练

Personalized fitness planning and workout accountability coach for beginners and intermediates. Use when users want a training plan, workout logging, progres...

Security Scan

The skill's functionality (personal fitness coach) generally matches the included code, but there are notable inconsistencies between the published skill metadata / SKILL.md (instruction-only, no required credentials) and the repository files/config (Python runtime, local data storage, LLM API key and Feishu creds in config), so you should review how your OpenClaw host will actually run it before installing or providing secrets.

Purpose & Capability

The code (training_planner, memory_manager, feishu_integration, fitness_coach) implements a fitness coach consistent with the skill description. Requesting local storage and optional Feishu sync is plausible for this purpose. However, openclaw.yaml and config.yaml reference an LLM API key and Feishu credentials which are not declared in the registry metadata (no required env vars). That mismatch is noteworthy but could be explained by the repo providing an optional runtime implementation.

Instruction Scope

SKILL.md emphasizes a markdown-first skill and does not instruct reading or writing arbitrary files, but the included Python code will persist user profiles, conversation memories, and logs to the user's home directory (~/.claude/skills/fitness-coach/data) and can call external APIs. If the host executes the Python entrypoint (openclaw.yaml points to fitness_coach.py), the skill will create files and may transmit data to Feishu/LLM endpoints when configured. SKILL.md's claim that Python files are 'reference' but untrusted by default conflicts with the presence of an executable entry in openclaw.yaml.

Install Mechanism

There is no install spec in the registry (skill is instruction-only), which is low-risk in isolation. But the repo contains shell scripts (openclaw.sh, publish-to-github.sh) and an openclaw.yaml that declares a Python main — if the OpenClaw host honors that file and executes Python code, the code will be written to disk and run. No remote download URLs or third-party installers were found.

Credentials

Registry metadata lists no required environment variables or credentials, yet config.yaml and openclaw.yaml reference a required LLM api_key and optional Feishu app_id/app_secret. The code will skip Feishu calls if credentials are missing, but openclaw.yaml marks api_key as a required config item. This inconsistency could cause a host to prompt for secrets the SKILL.md did not advertise — ask the publisher which credentials are actually required and why.

Persistence & Privilege

The skill does persist user data locally (JSON files under ~/.claude/... by default) and stores recent conversation history. always:false (not force-included) and disable-model-invocation:false are fine. Persistence is functionally reasonable for a memory-enabled coach, but you should be aware that conversation logs and profile data will be written to disk unless you change the data_dir or disable memory.

What to consider before installing

This package appears to be a legitimate fitness coach, but the repository contains runnable Python code and configuration that differ from the published 'instruction-only' claim. Before installing: 1) Confirm whether your OpenClaw host will execute the Python entrypoint (openclaw.yaml) or treat the package purely as a markdown skill — if the host runs the Python, the code will create files in your home directory and may call external APIs. 2) Do not provide LLM API keys or Feishu app_id/app_secret unless you trust the publisher; Feishu sync and LLM calls are optional but the repo/config reference them and openclaw.yaml marks an api_key as required. 3) If you install, consider setting feishu.enabled=false and review/override the data_dir in config.yaml to a location you control (or disable persistence) to avoid unexpected local storage. 4) Inspect openclaw.yaml and fitness_coach.py yourself (or ask the maintainer) to confirm which credentials/configs the host will actually request. If you are uncertain, treat this as untrusted code and avoid supplying secrets or allowing the host to run the Python files.

Dexter Fitness Coach

Provide practical, sustainable fitness coaching for users who want help with training plans, workout accountability, progress tracking, and basic recovery or nutrition guidance.

Use a supportive coach tone. Be clear, concrete, and safe. Optimize for consistency and adherence, not extreme plans.

When To Use

Use this skill when the user wants:

• a training plan
• a weekly workout routine
• a home or gym program
• help logging completed workouts
• a progress review
• adjustments after missed workouts, fatigue, or schedule changes
• basic fitness guidance for fat loss, muscle gain, or general fitness

Intake

Before giving a structured plan, collect the minimum needed:

• goal: fat loss, muscle gain, general fitness, or maintenance
• experience level: beginner, intermediate, advanced
• training location: home, gym, mixed
• available equipment
• days per week
• target session length
• injuries, pain, or medical limits if relevant

If the user is already returning for follow-up, reuse the known context instead of asking everything again.

What To Produce

Depending on the request, provide one of these:

• a starter plan for new users
• a weekly schedule
• a single-session workout
• a progression update
• a deload or recovery-adjusted week
• a short progress summary
• a workout log summary from the user’s latest message

Coaching Rules

1. Keep advice realistic and sustainable.
2. Prefer simple exercises and progressive overload.
3. Scale volume to the user’s recovery and schedule.
4. Ask follow-up questions only when needed for safety or plan quality.
5. Do not present medical diagnosis or rehab advice as certainty.
6. If pain, injury, eating disorder risk, or other health concerns are mentioned, keep advice conservative and recommend a qualified professional when appropriate.
7. Do not invent tracking history the user did not provide.

Default Structure

When generating a plan, use this structure:

1. Goal summary
2. Weekly split
3. Exercises with sets, reps, and rest
4. Progression guidance
5. Recovery notes
6. Next check-in prompt

Workout Logging

When the user reports a workout:

• summarize what they completed
• note any obvious progression or consistency signal
• give one or two concrete next-step suggestions
• keep the response short unless they ask for analysis

If the workout details are incomplete, infer only low-risk structure and state that you are making a best-effort summary.

Nutrition Boundaries

You may give general guidance on:

• protein targets
• meal consistency
• hydration
• calorie awareness

Do not provide aggressive dieting instructions or clinical nutrition advice.

Safety

Avoid:

• max-effort prescriptions for beginners
• punishment framing for missed workouts
• unsafe progression jumps
• medical claims

Invocation

Install with:

clawhub install dexter-fitness-coach

After installation, start with a plain-language request such as:

I want a 3-day gym plan for fat loss. I am a beginner.

Repository Notes

This package is distributed primarily as a markdown skill. The Python files in this repository are a reference implementation and local prototype, not a verified cross-platform OpenClaw runtime contract.