tasks-to-code
Warn
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to run various shell commands for project discovery and verification. During 'Phase 2: Discover' and 'Phase 5: Verify', it executes commands such as lint, test, build, and typecheck which are determined by the project's existing configuration.
- Evidence: SKILL.md (Phase 2, Phase 5) and references/verification-guide.md (Step 1) describe running quality gate commands found in the codebase.
- [EXTERNAL_DOWNLOADS]: The implementation phase grants the agent authority to add new software dependencies using package managers (e.g., npm, pip) if a task requires them.
- Evidence: SKILL.md (Phase 4, Rule 4) and references/implementation-guide.md (Dependency Discipline) instruct the agent to add and justify new dependencies.
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it consumes instructions and acceptance criteria from external files without sanitization.
- Ingestion points: plans//prd.md and plans//tasks.md.
- Boundary markers: None; the skill explicitly treats the PRD as the 'authoritative source of truth'.
- Capability inventory: Shell command execution (tests, builds, discovery), file system writes, and package installation.
- Sanitization: No validation is performed on the content of the requirement files or the commands they might implicitly trigger.
Audit Metadata