Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.
Community Detections
This directory holds detection rules contributed by the AiSOC community.
Anything you drop here as <your-rule>.yaml will be picked up by
scripts/build_marketplace.py on the
next pnpm marketplace:build and shown in the in-app marketplace with a
Community badge.
Authoring a community detection
Use any rule under detections/cloud/,
detections/identity/, or
detections/endpoint/ as a template. Every rule must:
- Declare a stable
id(kebab-case, no spaces). The convention for community contributions iscommunity-<github-handle>-<short-name>. - Map to MITRE ATT&CK via
tagsentries of the formmitre.attack.T1234ormitre.attack.T1234.567. The marketplace MITRE filter only sees rules with at least one such tag. - Set
severityto one oflow,medium,high,critical. - Reference a fixture under
detections/fixtures/so the detection can be replay-tested. - Include a short
false_positives:list (even if it is just[]) so responders know what to expect.
Validation
Before opening a PR:
python3 scripts/validate_detections.py
pnpm marketplace:build
pnpm marketplace:check
The marketplace:check step asserts that marketplace/index.json and its
public copy at apps/web/public/marketplace/index.json match what the build
script produces from the current contents of this repo. CI will fail
otherwise.
Promotion path
Community detections start with verified: false and source: "community".
After a maintainer reviews the rule against several weeks of real telemetry,
the rule may be promoted into the appropriate top-level category folder
(e.g. detections/identity/) and re-marked as verified.