doc-ctr-autopilot

Pass

Audited by Gen Agent Trust Hub on Mar 2, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes basic shell commands and local scripts for workflow management.
  • Evidence includes ls docs/08_CTR/CTR-{NN}_*/ for file existence checks during document detection.
  • The workflow executes a local Python script ai_dev_flow/scripts/update_traceability_matrix.py to maintain project documentation integrity.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external documentation content.
  • Ingestion points: Reads requirement documents from docs/07_REQ/REQ-*.md for contract generation.
  • Boundary markers: None explicitly defined to separate untrusted content from instruction logic.
  • Capability inventory: Performs file system reads and writes in docs/ directories, and executes subprocesses via shell and Python.
  • Sanitization: No explicit sanitization or validation of input document content is described prior to processing.
  • [DATA_EXFILTRATION]: File access patterns are restricted to specific project documentation paths (docs/07_REQ/, docs/08_CTR/, governance/plans/). There is no evidence of network-based exfiltration or access to sensitive system paths.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 2, 2026, 10:28 PM