project-init
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs state assessment and verification by executing local Node.js scripts.\n
- Evidence: Several
node -ecommands inSKILL.md(e.g., Phase 0 and Phase 4) execute logic from scripts located in./.claude/hooks/lib/.\n- [PROMPT_INJECTION]: The skill ingests data from project files (documentation, specs, and config) to drive its workflow, creating an indirect prompt injection surface.\n - Ingestion points: The skill is instructed to read
docs/project-config.json,docs/project-reference/docs-index-reference.md, and Feature Specs indocs/specs/.\n - Capability inventory: The skill can execute shell commands (git, node) and spawn sub-agents based on the state discovered in these files.\n
- Boundary markers: The instructions specify explicit routing for loading documents but lack mention of sanitization for the content processed from these external files.\n
- Sanitization: No explicit sanitization or filtering of external content is described.
Audit Metadata