feature-manager

Pass

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from external GitHub sources, which creates a potential surface for indirect instructions.
  • Ingestion points: The discover-feature.sh script reads GitHub issue bodies, and feature-feedback.sh fetches comment and review bodies from issues and PRs.
  • Boundary markers: While the skill uses specific markers for syncing issue bodies, it does not consistently use delimiters to wrap all external data shown to the agent.
  • Capability inventory: The skill possesses capabilities to perform git operations (git checkout, git merge, git push in cascade-update.sh) and modify GitHub issues (gh issue edit in the Sync Issue chore).
  • Sanitization: The skill mitigates risks by requiring explicit user confirmation before writing registry files or performing git merges, and routes all outgoing GitHub modifications through a post-review-gate for final oversight.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 9, 2026, 07:43 AM