videodb-monitoring

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to ask the user for their VideoDB API key and then embed that key verbatim into a command-line call (openclaw config set 'VIDEODB_API_KEY' 'sk-xxx'), which exposes secrets in LLM output and command arguments.

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

• Third-party content exposure detected (high risk: 0.80). The skill records and indexes the agent's screen (see SKILL.md example "Open example.com") and videodb.ts's search/summary/transcript flows explicitly ingest and return extracted visual scene text and audio transcripts (e.g., scene.text, segments) derived from arbitrary webpages or other third‑party content displayed on the screen, which the agent is expected to read and can materially influence subsequent actions.