issue-to-prd
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from GitHub issue titles, descriptions, and comments. This data is interpolated into a PRD draft that is published back to the GitHub issue.
- Ingestion points: Raw issue data is fetched using the GitHub CLI in stages/01-read-issue/CONTEXT.md.
- Boundary markers: The generated PRD uses markdown headers but lacks explicit delimiters or instructions to treat interpolated content as untrusted.
- Capability inventory: The skill can modify GitHub issues (gh issue edit), trigger remote missions (pylot dispatch), and perform network requests (curl).
- Sanitization: No explicit sanitization or filtering of the ingested content is documented.
- [COMMAND_EXECUTION]: The skill extensively uses the GitHub CLI (gh) to interact with repositories and a custom vendor CLI (pylot) to manage teams and dispatch worker missions.
- [EXTERNAL_DOWNLOADS]: The skill defines a mission brief that instructs a secondary worker agent to install developer dependencies (playwright and chromium) from the standard NPM registry.
- [DYNAMIC_EXECUTION]: The skill generates a mission brief containing shell commands and Playwright scripts based on the shared/prototype-mission-brief.md template. This brief is populated with specific repository and issue identifiers before being passed to a remote worker for execution.
- Execution method: The commands are executed on a spawned worker devbox to perform automated UI screenshotting tasks.
Audit Metadata