amendment-history
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user-supplied legal documents (PDFs or text) without explicit instruction isolation or boundary markers. An attacker could embed instructions within a contract file to attempt to manipulate the agent's output.
- Ingestion points: Document content is loaded via file uploads, pasted text, or planned future integrations (SKILL.md).
- Boundary markers: Absent. The instructions do not define delimiters (like XML tags) to separate document content from the agent's instructions.
- Capability inventory: Primarily text summarization and tracking; it lacks high-risk capabilities such as arbitrary command execution or network access.
- Sanitization: No explicit sanitization or validation of the document content is specified.
- [NO_CODE]: The skill consists entirely of natural language instructions in markdown format. It does not include any Python scripts, JavaScript files, or other executable binaries, which significantly reduces the technical attack surface.
Audit Metadata