coach
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEDATA_EXFILTRATIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill reads from and writes to hidden local files (~/.claude/coaching-diary.md and ~/.claude/coaching-resources.md) to manage session history and strategic insights. While these files are used to maintain context, they involve the storage and retrieval of personal strategic data within hidden filesystem paths.\n- [PERSISTENCE]: The skill implements a 'Session Memory Protocol' that appends session summaries, insights, and decisions to a local file at the end of every interaction. This is designed to provide continuity across different coaching sessions by parsing the diary at the start of each activation.\n- [INDIRECT_PROMPT_INJECTION]: The skill utilizes 'Web research' and 'User-sourced research' to develop domain expertise, which involves ingesting external, untrusted content into the agent's context.\n
- Ingestion points: External web search results and user-provided knowledge archives.\n
- Boundary markers: No specific delimiters or boundary warnings are provided to separate external data from system instructions.\n
- Capability inventory: The skill possesses read/write file access and web browsing capabilities.\n
- Sanitization: The instructions do not include protocols for sanitizing or validating information retrieved from the web before it is used to inform coaching guidance.
Audit Metadata