pr-fix

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill prompt instructs the agent to read an authentication token from a local file (cat ~/.config/forgejo/token) and then pass it directly in an Authorization header in a curl command (-H "Authorization: token $TOKEN"), which exposes the secret value to the LLM context and potentially its output.

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

• Third-party content exposure detected (medium risk: 0.30). The skill reads GitHub pull request reviews, comments, and CI details via the gh CLI, which exposes the agent to outsider-authored free text from PR contributors and reviewers.