auto-updater

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The workflow requires emitting full unified diffs (including hooks/hooks.json and .mcp.json) verbatim, which would force the agent to include any embedded API keys, tokens, or secrets present in those files in its output.

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

• Third-party content exposure detected (high risk: 0.85). The workflow fetches and diffs outsider-authored SKILL.md / source-registry content from installed skills’ upstream repositories at runtime (Step 1–2), and then ingests that fetched text into the agent context to generate unified diffs and run the skills-qa rescan (Step 2.5), creating an indirect prompt-injection path from untrusted repository text.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches commits and full diffs from source registries at runtime (e.g., git repository URLs / raw file URLs such as https://github.com/... or https://raw.githubusercontent.com/...), and those fetched diffs (including hooks/hooks.json and .mcp.json) are incorporated into prompts and can expose executable hooks, so external content directly controls the agent's prompt/context.