grant-proposal
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests content from local research specifications to generate proposal drafts.
- Ingestion points: Reads research specifications from quality_reports/specs/research_spec_*.md (SKILL.md).
- Boundary markers: Employs specific placeholders like [CLARIFY:] and [DELEGATE:] to manage external data mapping.
- Capability inventory: Uses Read, Write, Grep, Glob, Agent, and Task tools for file operations and skill coordination.
- Sanitization: Includes explicit instructions to follow .claude/rules/confidential-data.md and avoid promising the release of confidential data.
- [SAFE]: The skill is a text-based drafting tool that coordinates between other local skills and does not perform network operations, execute arbitrary code, or request elevated privileges.
Audit Metadata