fusion-github-review-resolution

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a legitimate workflow for resolving GitHub pull request threads. It uses standard developer tools (gh, git, jq) for its operations and restricts its network activity to official GitHub APIs.
  • [SAFE]: Helper scripts (get-review-comments.sh, resolve-review-comments.sh) include robust input validation for repository names and owner names using regular expressions to prevent common command injection or path traversal attacks.
  • [SAFE]: The resolve-review-comments.sh script implements a dry-run-first safety model and incorporates duplicate-reply guards to prevent the agent from repeatedly posting the same comment during execution retries.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes review comment bodies from pull requests.
  • Ingestion points: PR comment content is fetched via GraphQL in scripts/get-review-comments.sh.
  • Boundary markers: The skill does not employ technical delimiters (like XML tags) for external content, but it requires a structured reasoning step where the agent must 'judge each comment' before execution.
  • Capability inventory: The skill can modify local source code, create git commits, push to remote branches, and mutate GitHub thread states (reply and resolve).
  • Sanitization: No automatic sanitization of comment text is performed; the risk is mitigated by explicit instructions to not treat reviewers as automatically correct and to verify fixes against repository requirements.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 01:18 AM