empirical-proof

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to identify and execute shell commands resolved from the local project environment (e.g., npm test, npx jest, or custom scripts in CONTRIBUTING.md). While this is standard for coding agents, it introduces a reliance on the integrity of the project's own documentation.
  • [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it directs the agent to ingest and act upon instructions found in potentially untrusted local files.
  • Ingestion points: Project contributor documentation (e.g., CONTRIBUTING.md, AGENTS.md) and manifest files (e.g., package.json scripts block).
  • Boundary markers: None. The agent is told to "resolve the actual commands" without explicit safety wrappers for the extracted data.
  • Capability inventory: Execution of arbitrary shell commands as resolved from the repository content.
  • Sanitization: None specified for the resolved command strings before execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 06:02 PM
Security Audit — agent-trust-hub — empirical-proof