subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the execution of implementation plans by reading task descriptions and passing them to implementer and reviewer subagents. This external data is interpolated into the prompts defined in
implementer-prompt.mdandspec-reviewer-prompt.mdwithout the use of boundary markers or 'ignore instructions' warnings, which could allow instructions embedded within a task description to influence subagent behavior. - Ingestion points: Implementation plans and task text are read in
SKILL.mdand passed to subagents. - Boundary markers: Absent; task text is directly pasted into the prompt templates.
- Capability inventory: Subagents are instructed to implement code, write tests, and commit changes to the repository.
- Sanitization: No sanitization or validation of the plan text is performed before processing.
Audit Metadata