perps-review-pr

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the installation of CLI tools from the official npm registry (@anthropic-ai/claude-code, @openai/codex) and vendor sites (cursor.com). It recommends security measures like version pinning and manual verification of installers.- [COMMAND_EXECUTION]: Shell commands are used for managing git worktrees and running project validation scripts such as yarn workspace build and test. These are executed within an isolated worktree.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted git diffs and commit messages, creating an attack surface.
  • Ingestion points: PR branch data and file contents in skill.md.
  • Boundary markers: Prompt instructions to treat data as non-executable are present.
  • Capability inventory: File system access via git and execution of project validation scripts.
  • Sanitization: Relies on prompt-level instructions rather than automated filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 07:59 PM
Security Audit — agent-trust-hub — perps-review-pr