enforce-mantine-hooks
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes first-party source code to identify replacement candidates. Since source code is untrusted data that may contain malicious instructions in comments, this creates a surface for indirect prompt injection attacks against the orchestrator or parallel subagents.
- Ingestion points: Workflow Step 1 in SKILL.md (scanning first-party code) and the candidate list processing in worker-brief.md.
- Boundary markers: Absent. The instructions do not specify delimiters or techniques to distinguish code content from instructions for the LLM.
- Capability inventory: The skill can delete files (SKILL.md Step 4), modify source files (worker-brief.md Step 5), and execute shell commands (SKILL.md Step 5).
- Sanitization: Absent. Source code is processed directly as text for pattern matching.
- [COMMAND_EXECUTION]: The skill instructions direct the agent to run verification commands including typecheck, lint, and existing tests. These are standard commands for code refactoring but involve executing shell commands within the environment.
- [EXTERNAL_DOWNLOADS]: The skill references external documentation from https://mantine.dev/hooks/ as a fallback for the hook catalog. This is a reference to a well-known technology service for documentation purposes.
Audit Metadata