ai-implement
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content such as source code, commit history, and implementation plans, creating an attack surface for indirect prompt injection.
- Ingestion points: Context files (README.md, AGENTS.md) and feature-specific files (plan.md, research.md) read during execution.
- Boundary markers: The skill uses a dedicated 'Untrusted Content' section to instruct the agent to ignore imperatives in read data and treat them as findings.
- Capability inventory: The agent has access to the Edit tool for file modification, the Agent tool for sub-agent management, and shell execution for running tests.
- Sanitization: The skill relies on procedural reporting for detected imperatives in content rather than automated filtering.
- [COMMAND_EXECUTION]: The skill executes shell commands to retrieve the system date and run project-specific verification tests.
Audit Metadata