dt-obs-log-semantic-mapping

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

• Third-party content exposure detected (medium risk: 0.30). Workflow B1 reads only user-pasted content (which is outsider-authored free text) and then parses/deserializes it for validation, including running content-field analysis and proposing OpenPipeline transforms from that text.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I scanned the skill docs and sample files for literal, high-entropy credentials. I flagged values only where the literal looks like a session token / API token (long, random-looking string) embedded in the sample log content.

Findings:

• internal_session_id "00000bWt7w1aE4Ojc_P_Zm31x57RLHtLs_3z7_Ztd981" — long, random-looking session identifier inside a sample log content string; could be a reusable session token and therefore sensitive.
• rootApiTokenId / requestApiTokenId "00T3u94nuvXTk7aHJ697" — appears inside a transaction.detail object in a sample log content string; resembles an API-token identifier (and may be sensitive if it maps to a usable token).

Ignored items (not flagged) and why:

• "x-lbg-client-secret": "xxx" — placeholder / obvious example.
• Empty or redacted API key fields (e.g., apiKey:"") and short GUIDs/UUIDs, commit SHAs, request IDs, DT hashes — these are identifiers or example values rather than high-entropy credentials that grant access, so per policy they are not flagged.