Official Rules Library · signed & versioned

Governance rules for every AI agent

Discover, fork, and install declarative .rules that enforce policy on agent tool-calls before they run. Zero‑latency, zero‑tokens. Rulesets are authored by SigmaShake; community contributions welcome.

233
Public rulesets
2.3k
Rules enforced
14
Public plugins
Find what you need

Explore by category

More than just rules

Plugins: signed dashboard extensions

Browse plugins

Rule sets

Declarative .rules files that enforce policy on agent tool calls. Content-hash locked, Ed25519-signed by the author, and replayable in the playground. Install with ssg hub pull rules-typescript.

Plugins

Signed tarballs that extend the SSG dashboard UI. Workspace, observability, productivity panels, and more — installable on demand with ssg plugins install. Browse the marketplace.

Fresh from the marketplace

Recently added plugins

See more
How it works

From .rules to enforced policy in three steps

  1. 01

    Discover or fork

    Browse signed, versioned rulesets authored by SigmaShake — and submit your own. Each ruleset is fetched directly from GitHub and locked to a content hash.

    233 public rulesets
  2. 02

    Install with one command

    Run ssg hub pull rules-typescript and the engine pulls, verifies the signature, and writes the ruleset into your project's .sigmashake/ directory.

    Works with Claude, Cursor, Copilot, Codex
  3. 03

    Enforce at tool-call boundary

    Every agent tool call is evaluated against the ruleset before execution — zero-latency, zero-tokens. Decisions: ALLOW, DENY, ASK, FORCE, LOG, SHADOW; FORCE returns a safe substitute path.

    Hooks-based · no proxy · no LLM in the loop
Latest releases

Recently updated

See more
Email
@sigmashakeinc/rules-email
Official Verified
Gmail and Microsoft Outlook governance for AI assistants. These rules keep an agent from sending, deleting, or auto-forwarding email on your behalf without a human in the loop — the controls a non-technical user wants when Claude Cowork or the Codex desktop app is connected to their mailbox over MCP.
sigmashakeinc v1 · 2 days ago 4 rules
Docs
@sigmashakeinc/rules-docs
Official Verified
Google Docs, Microsoft Word, and Google Drive governance for AI assistants. These rules block public ('anyone with the link') sharing and require approval before an agent deletes a document or changes who can access it — the controls a non-technical user wants when an AI app is connected to their documents over MCP.
sigmashakeinc v1 · 2 days ago 4 rules
Common
@sigmashakeinc/rules-common
Official Verified
sigmashakeinc v5 · 2 days ago 30 rules
Cloud
@sigmashakeinc/rules-cloud
Official Verified
sigmashakeinc v5 · 2 days ago 12 rules
Cicd
@sigmashakeinc/rules-cicd
Official Verified
sigmashakeinc v5 · 2 days ago 5 rules
Backend
@sigmashakeinc/rules-backend
Official Verified
sigmashakeinc v5 · 2 days ago 7 rules
Amp
@sigmashakeinc/rules-amp
Official Verified
sigmashakeinc v2 · 2 days ago 3 rules
Aider
@sigmashakeinc/rules-aider
Official Verified
sigmashakeinc v5 · 2 days ago 6 rules
Ai-ml
@sigmashakeinc/rules-ai-ml
Official Verified
sigmashakeinc v5 · 2 days ago 7 rules
Windsurf
@sigmashakeinc/rules-windsurf
Official Verified
sigmashakeinc v3 · 2 days ago 6 rules
Slack
@sigmashakeinc/rules-slack
Official Verified
Slack governance for AI assistants. These rules require human approval before an agent posts a message or pings a whole channel (@channel / @here / @everyone) and keep an audit trail of channel-history reads — the controls a non-technical user wants when an AI app is connected to their Slack over MCP.
sigmashakeinc v1 · 2 days ago 3 rules
Sheets
@sigmashakeinc/rules-sheets
Official Verified
Google Sheets and Microsoft Excel governance for AI assistants. These rules block spreadsheet formulas that exfiltrate data to external URLs (IMPORTXML / IMPORTDATA / IMPORTRANGE / WEBSERVICE) and require approval before an agent deletes a sheet or clears cells — the controls a non-technical user wants when an AI app is connected to their spreadsheets over MCP.
sigmashakeinc v1 · 2 days ago 3 rules
Trust & integrity

Engineered for auditors

Trust center

Signed releases

Every published ruleset is signed with the maintainer's key. The ssg CLI verifies the signature against /api/pubkey before installing.

Content-hash locked

Each version pins a SHA-256 of the indexed rules. Any edit upstream bumps the hash and the version — no silent overrides.

Transparency log

Every publish, update, and resync appends to an append-only Merkle log cross-anchored with sigmashake-compliance.

MCP-native

Connect any agent to mcp.sigmashake.com/sse — six tools cover eval, search, audit, and publish. No SDK required.

Security evidence program

Internal control evidence is collected continuously via sigmashake-compliance with 90-day Object Lock retention. Not a completed SOC 2 attestation.

Contribute a ruleset

Publish rules in one command

Got a ruleset that would protect other teams? Every submitted ruleset is signed, versioned, and auditable. Install ssg, point it at your repo, and it's live on the hub — instantly discoverable by every AI agent host.

~/your-project
$ curl -fsSL https://install.sigmashake.com | sh
$ ssg publish
✓ Parsed 14 rules across 3 technologies
✓ Signed & pushed to hub.sigmashake.com
$ ▍