progress
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use shell commands for repository management and validation, including 'git status', 'git worktree list', 'git commit', 'git push', and 'npm test'. These are used to reconcile project state and validate implementation claims.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted instructions from repository-local files. Ingestion points: reads 'AGENTS.md', 'ROADMAP.md', 'PLAN.md', and various project items in 'plans/' and 'tasks/' directories. Boundary markers: implements 'Goal Contracts' and 'Execution Status' as logical boundaries; instructions emphasize reconciling repository data with direct user instructions. Capability inventory: identifies the ability to modify local files, execute git operations, and run validation scripts. Sanitization: requires the agent to verify material claims against repository evidence rather than trusting document content alone.
Audit Metadata