log-leave
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data and user input to update a local register, creating an injection surface.\n
- Ingestion points: The skill ingests data from the user regarding employee details (Step 2) and from a local markdown file
~/.claude/plugins/config/claude-for-legal-zh/employment-legal/CLAUDE.md(Step 1).\n - Boundary markers: Absent. The instructions do not define delimiters or provide warnings to the agent to disregard instructions potentially embedded within the ingested data.\n
- Capability inventory: The skill performs file reading (Step 1) and file writing (Step 5) within the local plugin configuration directory.\n
- Sanitization: Absent. There is no instruction to validate, filter, or escape user-supplied strings before they are formatted as YAML and written to the filesystem, which could lead to schema confusion or persistence of malicious instructions if the register is processed by other skills later.
Audit Metadata