spec-driven-develop

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs the agent to query and read GitHub resources (issues, issue bodies/comments, and milestone descriptions) as part of its required workflow (see references/github-integration.md and Phase 4/Adaptive Control where it reads/writes Milestone description YAML and Issue telemetry), and those GitHub fields are user-generated/untrusted content that directly influence task execution and control decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly performs runtime GitHub API calls (e.g., reading/updating milestone descriptions via gh api and referencing the workflow repo) — for example https://github.com/zhu1090093659/spec-driven-develop — and it relies on YAML stored in those milestones/Master.md to drive adaptive control (halt/replan/rescope) and agent decisions, so the external GitHub content directly controls agent behavior at runtime.