pdd-feature
Pass
Audited by Gen Agent Trust Hub on Nov 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from issue descriptions and user replies while maintaining capabilities for file modification and command execution. 1. Ingestion points: The agent reads external data from issue descriptions (understand.md), repository rules, and user interaction captured via the workflow approve command. 2. Boundary markers: While the skill employs safe shell practices like quoted heredoc delimiters (<<'PDD_SAID') to record user feedback, it lacks explicit natural language boundary markers or instructions to disregard potential commands within processed issue text. 3. Capability inventory: The agent utilizes a suite of development tools (pdd files, pdd lint, pdd test, pdd simulate) and is instructed to execute code paths under development for review. 4. Sanitization: There are no explicit instructions for the agent to sanitize or validate the content of external inputs before they are incorporated into prompts or used in execution steps.
- [COMMAND_EXECUTION]: The skill executes local development utilities (pdd) and environment-specific tools as part of its feature development workflow. This behavior is consistent with the skill's intended purpose as a development assistant.
- [DYNAMIC_EXECUTION]: The skill is instructed in the review phase to run the actual changed code paths to verify behavior and compare the results against expected boundaries or mocks.
Audit Metadata