find-simplifications
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted repository content (source code, architecture documents, and pull request diffs) to generate design proposals called Agent Notes.
- Ingestion points: SKILL.md and references/folding-prs.md explicitly instruct the agent to read AGENTS.md, architecture docs, and production source code.
- Capability inventory: The agent has the ability to search files using rg, create or modify files in the .agents/notes/ directory, and execute local project scripts such as pnpm run lint and pnpm run doc-sync.
- Boundary markers: There are no explicit instructions to use delimiters or ignore instructions potentially embedded within the ingested code or documentation.
- Sanitization: The instructions do not provide requirements for sanitizing or escaping content extracted from the repository before it is used in note generation.
- [COMMAND_EXECUTION]: The skill makes use of local command-line tools and repository-specific scripts to perform its auditing and validation tasks.
- Evidence: SKILL.md describes using rg for symbol reachability checks and suggests running validation commands like pnpm run doc-sync and pnpm run lint. These are standard development operations scoped to the local environment.
Audit Metadata