fusion-rules
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the repository which could be used to influence the agent's behavior during rule generation.
- Ingestion points: The agent scans repository documentation and configuration files (README, CONTRIBUTING, ADRs, CI workflows) in Step 2 of the workflow defined in
claude-code.agent.md,copilot.agent.md, andcursor.agent.md. - Boundary markers: The
SKILL.mdfile contains explicit instructions in the Safety section to "Show drafts before writing any files" and "No overwrites without showing diff and getting approval." - Capability inventory: The skill is designed to write instruction files to the local filesystem (e.g.,
CLAUDE.md,.cursor/rules/*.mdc,.github/copilot-instructions.md). - Sanitization: The workflow relies on the user's manual review of the generated drafts and diffs rather than automated content filtering.
Audit Metadata