team-code-review

Pass

Audited by Gen Agent Trust Hub on Nov 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from the repository and external agent reports, using this data to steer the behavior of an 'implementer' agent.
  • Ingestion points: The skill reads 1-task.md, 8-plan.md, and the output reports from the code-reviewer, technical-writer, ux-reviewer, and verifier agents.
  • Boundary markers: No explicit delimiters or instruction-ignore warnings are specified for the payload sent to the implementer agent.
  • Capability inventory: The skill executes git commands, local shell/Node.js scripts, and performs file system writes for review records and the TodoWrite ledger.
  • Sanitization: Reviewer findings are aggregated, joined by spaces, and relayed to the implementer without escaping or sanitization of potential injection content.
  • [COMMAND_EXECUTION]: The skill relies on the execution of several internal scripts and git commands to analyze the environment and manage the review state.
  • Executes scripts/change-profile.mjs via node to determine review depth based on git diff metrics.
  • Executes scripts/tree-check.sh and discover-topic.sh via the shell to monitor worktree integrity and resolve artifact paths.
  • Frequently invokes git with the -C parameter to perform status, diff, and rev-parse operations across various worktrees.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 15, 2026, 12:30 PM