night-market-change-control
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to process and review untrusted content from Pull Request diffs, PR descriptions, and Architecture Decision Records (ADRs). This ingestion process creates a potential surface for indirect prompt injection where malicious instructions embedded in a contribution could attempt to manipulate the agent's review logic or bypass safety rules. \n
- Ingestion points: Pull Request diffs, PR descriptions, and markdown files in
docs/adr/. \n - Boundary markers: The skill defines a strict hierarchy of project rules (Constitution, ADRs) and pre-commit hooks, although it does not document specific delimiters used to isolate LLM instructions from PR data. \n
- Capability inventory: The skill utilizes git for branching and tagging, executes local testing suites via
make test, and runs static security scanners. \n - Sanitization: It includes active security measures such as
banditscanning, identity leak detection (slop-detector), and validation against package name hallucinations in AI-suggested dependencies.
Audit Metadata