git-commit

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly requires the agent to "主动查阅" and use Read/WebFetch to fetch the public GitHub raw file (https://raw.githubusercontent.com/.../commit-types.ts) to obtain emoji/type definitions, a public user-controlled resource the agent must read and act on to construct commit messages, which could carry malicious instructions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent to "主动查阅" (use Read/WebFetch at runtime) to fetch definitions from https://raw.githubusercontent.com/ruan-cat/monorepo/dev/configs-package/commitlint-config/src/commit-types.ts, and that fetched content directly controls the emoji/type rules used to generate commit messages (agent instructions), so this is a runtime external dependency that controls prompts.