session-search

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly reads session transcripts and prints raw user turns (the python snippet prints c[:400]), which can cause any secrets present in those messages (API keys, tokens, passwords) to be emitted verbatim.

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md shows the runtime path obsidian-wiki sessions-query "<topic>" followed by sessions-show <session-id> --pretty and then LLM-side filtering/reading of the returned transcript JSONL content; because the user supplies the topic (free text) that can cause the workflow to fetch and ingest outsider-authored session transcripts, this exposes the agent to indirect prompt injection.