write-feature-docs

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). SKILL.md requires reading spec files (specs/<id>/PRODUCT.md and specs/<id>/TECH.md) and then generating an MDX draft from them, so outsider-authored free text from those spec documents can be ingested into the agent’s LLM context at runtime.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly reads spec files and runs GitHub CLI/API commands against the repositories warpdotdev/warp-internal and warpdotdev/docs at runtime (e.g., "specs//PRODUCT.md", gh search code ... --repo warpdotdev/warp-internal, and gh api repos/warpdotdev/warp-internal/git/trees/HEAD?...) which fetches remote content that directly controls the agent's generated prompts and output (the spec content is injected into the drafting process).