feature-to-plan
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from GitHub issues or user-provided text to seed feature plans. It implements specific defenses to mitigate risks:
- Ingestion points: Phase 1 (Orient) in
SKILL.mdreads issue content via theread_issuetool or user-pasted text. - Boundary markers: The skill structures output using defined markdown headers and the EARS (Easy Approach to Requirements Syntax) format as defined in
references/spec-format.md. - Capability inventory: The skill utilizes file system tools (
Read,Write,Edit), shell access (Bash), and GitHub API interactions (mcp__github) to perform reads and create new issues. - Sanitization: As detailed in
references/github-output.md, instructions explicitly require treating seed text as data, converting@mentionsto inline code to avoid unwanted notifications, and removing issue-closing keywords like 'Fixes' or 'Closes' to prevent unintended side effects in the repository workflow.
Audit Metadata