subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines a legitimate development coordination workflow. It utilizes localized prompt templates to delegate tasks to subagents, ensuring isolation of context and history. No hardcoded credentials, malicious network operations, or obfuscated payloads were detected.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process implementation plans, which serves as a potential surface for indirect prompt injection if those plans contain malicious instructions.
- Ingestion points: The agent is instructed to read plan files (e.g.,
docs/superpowers/plans/feature-plan.md) and extract task text. - Boundary markers: The skill uses Markdown headers (e.g.,
## Task Description,## Context) to separate task content from subagent instructions, though it lacks strict delimiters or "ignore instructions" directives. - Capability inventory: Subagents are equipped with capabilities to write files, execute tests, and perform git operations (commits and branch management).
- Sanitization: Task content from external plans is interpolated directly into the
implementer-prompt.mdandspec-reviewer-prompt.mdtemplates without explicit sanitization or escaping.
Audit Metadata