apply-rules

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly fetches and decodes rule files from arbitrary GitHub URLs as the "source" (see "Step 2: Fetch Source Rules" which uses gh api to list and download .md files), and it parses and applies that untrusted, user-provided content to drive detection, merging, cleanup, and file-write decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches rule files at runtime from a GitHub source (e.g. https://github.com/org/repo/tree/main/.claude/rules and via gh api repos/{owner}/{repo}/contents/{path}?ref={branch}), decodes and injects those .md files into the agent's processing flow so the external content directly controls merging/behavior—this meets the criteria for a runtime external dependency that controls agent prompts/decisions.