purge-private-vocab

Pass

Audited by Gen Agent Trust Hub on May 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill demonstrates an indirect prompt injection surface due to its core function of ingesting and processing untrusted text from external sources.\n
  • Ingestion points: Processes user-provided target documents (like PR descriptions or Jira tickets) and local plan files (located in ~/.claude/plans/**/*.md).\n
  • Boundary markers: Absent. The instructions do not define delimiters or specific 'ignore' directives to isolate the content of the processed documents from the agent's logic.\n
  • Capability inventory: The skill has the authority to read and edit local files using standard CLI tools.\n
  • Sanitization: Absent. The logic relies on heuristic pattern matching rather than rigorous validation or sanitization of the input text.\n- [SAFE]: No unauthorized network connections, credential harvesting, or malicious code execution patterns were found. The skill incorporates a 'Dry-run' report which acts as a safety checkpoint for users to verify changes before they are committed to disk.
Audit Metadata
Risk Level
SAFE
Analyzed
May 17, 2026, 09:35 AM
Security Audit — agent-trust-hub — purge-private-vocab