Only the latest release on main receives security fixes.

Do not open a public GitHub issue for security vulnerabilities.

Report privately via GitHub's Security Advisories feature.

Include:

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix if known

You will receive a response within 7 days. If the vulnerability is confirmed, a patch will be released as soon as possible with a coordinated disclosure.

This package is a read-only analysis library over records acquired by commoner-probe. It optionally makes outbound HTTP requests to a local Ollama endpoint for the LLM discourse tier. It writes JSONL and SQLite files locally.

Known constraints:

• All three LLM call sites — the discourse-tier classifier (analyse-discourse --llm-tier), the ministry-query refinement path, and the topic classifier's llm mode (--classifier llm / a topic profile's classifier.mode: "llm") — validate the endpoint scheme (HTTP/HTTPS only) via llm_client.py, and can optionally reject loopback/private/link-local hosts (allow_private=False; "allow_private": false in a topic profile's classifier block) to defeat SSRF against internal services.
• No authentication credentials are stored or transmitted beyond what the caller supplies
• data/, notes/, and memory/ directories are gitignored and must never be committed