investigation-add
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and process untrusted external data such as interview transcripts, emails, and evidence files. An attacker could embed malicious instructions within these documents to influence the agent's behavior during the evaluation or logging phase.
- Ingestion points: User-provided text and attachments supplied through the
argument-hintparameter inSKILL.md. - Boundary markers: Absent. The instructions do not define clear delimiters or "ignore" instructions to separate the input data from the prompt logic.
- Capability inventory: The skill reads a local configuration file (
~/.claude/plugins/config/claude-for-legal-zh/employment-legal/CLAUDE.md) and performs content evaluation based on defined criteria. - Sanitization: Absent. There is no mechanism described for filtering, escaping, or validating the input data before processing.
Audit Metadata