feature-video

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill reads and re-inserts text from web pages and PR bodies (SAVED_TEXTAREA / PR body) into generated eval commands and a gh CLI --body argument, which forces the agent to include user-provided text verbatim (potentially secrets/tokens) in its output.

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly navigates to and reads GitHub PR pages and metadata (via gh pr view and agent-browser opening https://github.com/[owner]/[repo]/pull/[number], reading the PR body and comment textarea) which are user-generated, third-party content that the agent parses and uses (e.g., extracting upload URLs and PR text), so untrusted page content could influence its actions.