load-project
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is purely instructional and defines a protocol for project context initialization. It contains no executable code, binaries, or external dependencies. The instructions explicitly mandate that higher-ranking security and authorization boundaries remain in force and that the agent must not perform broad, unrestricted file searches.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external project files, which constitutes a potential attack surface for indirect prompt injection. 1. Ingestion points: Project root documents (e.g., CLAUDE.md), workspace rules, and task-specific files. 2. Boundary markers: The instructions provide a specific rule hierarchy and role-based document selection to limit data ingestion. 3. Capability inventory: The skill relies on the underlying agent's existing file-reading capabilities and does not introduce new tools or permissions. 4. Sanitization: The protocol explicitly requires redacting secrets, personal data, and confidential content before generating the final status report.
Audit Metadata