panel-review

Pass

Audited by Gen Agent Trust Hub on May 11, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it fetches and processes untrusted content from GitHub, including PR titles, descriptions, and review comments.
  • Ingestion points: Data is gathered via gh pr view, gh pr diff, and gh api calls within the panel-review.sh script.
  • Boundary markers: The prompt templates (prompts/review-pr.md, prompts/review.md) use Markdown headers to delimit external data sections.
  • Capability inventory: In worktree mode, sub-agents are granted workspace-write or bypassPermissions to perform shell operations and run tests.
  • Sanitization: There is no automated sanitization of the fetched data; instead, the SKILL.md provides a detailed "Deep Mode" procedure for manual verification of agent findings against the source code.
  • [COMMAND_EXECUTION]: The panel-review.sh script executes several local CLI tools including git, gh, and multiple AI agent interfaces (codex, claude, opencode). It manages these as parallel subprocesses and provides them with varying levels of access to the local filesystem and shell, which is necessary for the skill's primary function of running builds and tests during a review.
Audit Metadata
Risk Level
SAFE
Analyzed
May 11, 2026, 09:15 PM
Security Audit — agent-trust-hub — panel-review