The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or data exfiltration vectors were detected. The skill is well-structured and incorporates extensive safety protocols for health and medical emergencies.
[COMMAND_EXECUTION]: The skill uses a bundled Python script (scripts/analyze_log.py) to analyze training logs and calculate estimated 1-rep max (1RM) trends, volume, and fatigue signals. The execution is limited to this local, static script using standard libraries and does not involve remote code or dynamic generation.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting data from profile.md and log.jsonl. However, the risk is assessed as safe because the data is maintained locally by the user, and the ingestion process involves canonical mapping of exercise names and structured JSON parsing.
Ingestion points: Training history is read from workout-coach/log.jsonl and user preferences from workout-coach/profile.md at the start of every session.
Boundary markers: The skill relies on the JSONL and Markdown file structures; it does not implement explicit "ignore embedded instructions" delimiters within the files.
Capability inventory: The skill possesses the ability to read and write files in the local workout-coach/ directory and execute the bundled Python analysis script.
Sanitization: User input is sanitized via a Hebrew-to-English canonical exercise map, and log data is parsed using a standard JSON library in the analysis script.
[EXTERNAL_DOWNLOADS]: The skill references several authoritative health and fitness websites (e.g., Israeli Ministry of Health, NASM, and emotional support hotlines). These are recognized as trusted, well-known services and are used solely for informational and safety routing purposes.

israeli-workout-coach