autodev-parallel
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes an external document (specified in the
prdfield, e.g.,PRD.md) to categorize tasks and generate instructions for sub-agents. This creates a surface for indirect prompt injection where malicious content in the PRD could influence sub-agent behavior. - Ingestion points: The PRD file content is read and its items are interpolated into prompts for parallel agents in Phase 2.
- Boundary markers: Absent. Item descriptions are used directly in sub-agent prompts without delimiters or warnings to ignore embedded instructions.
- Capability inventory: The orchestrator and its sub-agents have access to
Bash(for command execution and git operations),Write,Edit, andAgenttools. - Sanitization: No evidence of sanitization or validation of the PRD content before interpolation.
- [COMMAND_EXECUTION]: The skill executes a user-provided
verifycommand using theBashtool to validate the work performed by sub-agents. While a core part of its development workflow, this allows for the execution of arbitrary shell commands defined in the skill's configuration or the PRD file.
Audit Metadata