# SigmaShake Rules Hub > Public registry of governance rules for AI agents ## What is SigmaShake? SigmaShake is a governance engine that evaluates AI agent tool calls against declarative rules before execution. It can allow, deny, ask, force safe alternatives, log, or shadow-audit at the tool-call boundary with zero-token enforcement. ## This Service: Hub The Hub at hub.sigmashake.com hosts two distinct surfaces: - Rule sets — declarative .rules files that enforce policy on agent tool calls. Browse at / - Plugins — signed tarballs that extend the SSG dashboard UI. Browse at /plugins Both share the same Ed25519 signing root; rule sets carry per-author content signatures, plugins carry per-author artifact signatures. ## JSON API Endpoints - GET /api/rulesets?page=0&limit=20&tech=ts&q=keyword — Paginated rulesets listing - GET /api/rulesets/:id — Single ruleset with all rules and raw .rules syntax - GET /api/plugins?page=0&limit=20&category=tools&q=keyword — Paginated plugin marketplace listing - GET /api/plugins/:id — Plugin detail (manifest URL, SHA-256, Ed25519 signing fingerprint) - GET /api/plugins/categories — Per-category plugin counts - GET /api/search?q=keyword&tech=ts&format=json — Search rulesets (JSON) - GET /api/technologies — List all available technologies - POST /api/downloads/:id — Register a download (rate limited: 1/min/IP/ruleset) - POST /api/eval — Evaluate a tool call against a ruleset (no install required, rate limited: 10/min/IP/ruleset) - GET /api/health — Service health check ## Publishing Rules (AI Agent / Programmatic) POST /api/submit — Submit a ruleset without browser OAuth Headers: Authorization: Bearer Body: {"repo": "https://github.com/owner/repo", "path": ".sigmashake/rules"} Response: {"ok": true, "ruleset_id": "uuid", "hub_url": "https://hub.sigmashake.com/ruleset/uuid", "rule_count": 13, "technologies": ["ts"]} Token scope required: public_repo (read public repos) ## Quick Publish (ssg CLI) Install: pnpm add -g ssg Publish: ssg publish (creates GitHub repo + pushes rules + opens hub pre-filled) One command. Requires: GitHub CLI (gh) authenticated. ## MCP Server For AI agents using Model Context Protocol, connect to: https://mcp.sigmashake.com/sse Tools: sigmashake_eval, sigmashake_rules_list, sigmashake_rules_get, sigmashake_audit, sigmashake_status, sigmashake_hub_submit ## Documentation - Full docs: https://docs.sigmashake.com - Rule syntax: https://docs.sigmashake.com/rule-syntax - Getting started: https://docs.sigmashake.com/getting-started - Architecture: https://docs.sigmashake.com/architecture ## Rule Syntax (Quick Reference) ``` rule block-dangerous-commands { priority 100 severity error DENY execution IF command CONTAINS "rm -rf" OR command CONTAINS "git push --force" MESSAGE "Destructive command blocked." } ``` Fields: command, path, content, tool, input. Operators: CONTAINS, EQUALS, STARTS_WITH, ENDS_WITH, GLOB, REGEX, WORD, LINE_CONTAINS, LINE_REGEX Decisions: ALLOW (permit), DENY (runtime: block), ASK (human approval), FORCE (refuses the original risky action and returns a safe substitute or required safer path), LOG (allow+audit), SHADOW (silent audit) Targets: execution, read, write, edit, search, agent, network, any ## Contributing Submit rulesets via GitHub OAuth at /submit. Rules must be in a public GitHub repository. Source: https://github.com/sigmashakeinc/sigmashake-mono