port-feature

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's operational instructions are confined to discovery and documentation. It explicitly forbids code implementation, configuration edits, or bulk repository reading, limiting its impact to the creation of a single gap-map markdown file.
  • [INDIRECT_PROMPT_INJECTION]: The skill analyzes external codebases, which is a potential injection surface, but mitigates this via strict process rules.
  • Ingestion points: Reads Reference and Target codebases, CONTEXT.md, specs/adr/, and AGENTS.md to identify feature behaviors and target stack rules.
  • Boundary markers: Instructs the agent to link or cite evidence rather than pasting raw content into the artifact.
  • Capability inventory: Capabilities are restricted to targeted search (rg, git grep) and writing to a specific path within <artifacts-root>/specs/port/.
  • Sanitization: Output is constrained to a predefined nine-section template with a limit of three bullets per section, reducing the likelihood of processing or emitting malicious payloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 06:25 PM
Security Audit — agent-trust-hub — port-feature