guided-review

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data including Git diffs, worktree changes, and repository-specific instruction files.
  • Ingestion points: Git diffs (git diff), repository status (git status), and "nearest repository instructions" (e.g., .claudecode.md) located in SKILL.md.
  • Boundary markers: The instructions explicitly define the user-selected scope as the review boundary and caution against expanding it.
  • Capability inventory: The skill can stage and commit files, run shell commands for validation, and send data to external MCP tools.
  • Sanitization: The skill contains explicit instructions to redact secrets and unrelated content before sending data to external services.
  • [COMMAND_EXECUTION]: The skill uses various Git commands and shell-based validation tools to perform its tasks.
  • Evidence: Usage of git status, git diff --cached, git commit, and sha256sum for integrity checking.
  • [DYNAMIC_EXECUTION]: The skill executes repository-local tools and validation scripts based on instructions found within the analyzed repository.
  • Evidence: The skill is instructed to "Run the narrowest relevant syntax checks, tests, linters... required by the repository instructions" in SKILL.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 09:53 AM
Security Audit — agent-trust-hub — guided-review