laravel-eloquent
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill mandates a workflow where the agent must explore the codebase and query documentation before implementation. This ingestion of untrusted project data (e.g., existing PHP files, model metadata) creates a surface for indirect prompt injection if those files contain instructions disguised as comments or code.
- Ingestion points: The workflow requires using "fuse-ai-pilot:explore-codebase" to inspect existing models and "mcp__context7__query-docs" to query documentation patterns.
- Boundary markers: None specified; the skill does not provide clear delimiters or 'ignore embedded instructions' warnings when processing ingested project data.
- Capability inventory: The agent is instructed to implement code for models, observers, and factories, and then validate the implementation using "fuse-ai-pilot:sniper".
- Sanitization: No explicit sanitization or filtering of the content retrieved from the codebase is mentioned.
Audit Metadata