spec

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime path “Step 1 — Detect Source and Fetch Context” ingests outsider-authored free text from user-provided issue URLs by reading Linear issues (mcp__plugin_linear_linear__get_issue), GitHub issue bodies (gh issue view ... --json title,body,...), and Jira pages via WebFetch, before generating the spec.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches external issue content at runtime (e.g., https://github.com/*/issues/, https://.linear.app//issue/, https://.atlassian.net/browse/) and uses that content to drive/specify the agent's outputs, so remote content directly controls prompts.