Imported from githubmofo/TorusGuard (SKILL.md). Install upstream with npx skills add githubmofo/TorusGuard. Copyright stays with the author.

TorusGuard Master Security Router

Name: torusguard
Author: githubmofo

Principle: If the browser or client receives it, users can inspect it. Keep secrets, database queries, and authorization strictly on trusted server-side code.

⚡ Workspace Bootstrap Check

If .torusguard/ exists on disk: Full autonomous governance active. Load .torusguard/memory/context.json into prompt context for local-first intelligence. Route commands to specialist workflows below.
If .torusguard/ is absent (Autonomous AI Mode): Operates standalone in any AI agent (Kimi, Antigravity, VS Code, Cursor, Claude Code, Windsurf). Apply universal security invariants:
- Server-side auth & tenant scoping (.filter(tenant=...), .Where("tenant_id = ?"), where: { tenantId: ... }).
- Zero client-side secrets (SUPABASE_SERVICE_ROLE_KEY, live API keys, Stripe secret keys).
- Ponytail Protocol patch limits ($\le 35$ additions, $\le 25$ deletions, zero full-file rewrites).
- Upgrade: Run npx torusguard init to unlock local .torusguard/ workspace tracking, persistent memory, and individual slash commands.

Specialist Skill Routing (Lazy Loading)

Load only the designated sub-skill on demand to preserve the 1,000–1,500 token budget:

Command	Specialist Skill	Agent	Purpose
`/torusguard init`	`skills/torusguard-init/SKILL.md`	`profiler`	Workspace discovery, stack detection & scaffolding
`/torusguard authorize`	`skills/torusguard-authorize/SKILL.md`	`reviewer`	Legal scope definition & safety boundaries
`/torusguard audit`	`skills/torusguard-audit/SKILL.md`	`auditor`	Static AST security audit & root-cause clustering
`/torusguard verify`	`skills/torusguard-verify/SKILL.md`	`validator`	Evidence sufficiency & line match audit
`/torusguard web-validate`	`skills/torusguard-web-validate/SKILL.md`	`validator`	Authorized non-destructive HTTP probing
`/torusguard exploit-check`	`skills/torusguard-exploit-check/SKILL.md`	`validator`	Bounded single-step exploitability confirmation
`/torusguard harden`	`skills/torusguard-harden/SKILL.md`	`remediator`	Governed remediation under Ponytail Protocol
`/torusguard apply`	`skills/torusguard-apply/SKILL.md`	`remediator`	Governed patch application with rollback snapshots
`/torusguard recheck`	`skills/torusguard-recheck/SKILL.md`	`reviewer`	Differential AST re-scan & closure verification
`/torusguard report`	`skills/torusguard-report/SKILL.md`	`reviewer`	Executive posture reporting, SARIF & visual HTML export
`/torusguard status`	`skills/torusguard-status/SKILL.md`	`reviewer`	Diagnostic overview of posture, stack & rules
`/torusguard memory`	`skills/torusguard/SKILL.md`	`reviewer`	Persistent memory inspection, decay & context cards
`/torusguard full`	`skills/torusguard-full/SKILL.md`	All	End-to-end 7-stage closed-loop execution

Note: Never load all skills simultaneously. Lazy-load strictly on demand.

{ "ocm": "1", "id": "githubmofo-torusguard-torusguard", "kind": "skill", "name": "torusguard", "description": "Universal autonomous security engine: 71 rules across 11 families, polyglot stack detection across 16+ languages, governed remediation (Ponytail Protocol <=35 add, <=25 del), SARIF v2.1.0 exports, single-file HTML dashboards, and persistent security memory context for modern web applications.", "publisher": "githubmofo", "version": "1.0.0", "capabilities": { "domains": [ "math" ], "tags": [ "skill-md", "github" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "Universal autonomous security engine: 71 rules across 11 families, polyglot stack detection across 16+ languages, governed remediation (Ponytail Protocol <=35 add, <=25 del), SARIF v2.1.0 exports, single-file HTML dashboards, and persistent security memory context for modern web applications." ], "primary": false, "metadata": { "source": { "provider": "github", "repository": "https://github.com/githubmofo/TorusGuard", "path": "SKILL.md", "ref": "5d392aed7885d591559c76c42e5f999d5fdd0745", "url": "https://github.com/githubmofo/TorusGuard/blob/5d392aed7885d591559c76c42e5f999d5fdd0745/SKILL.md", "key": "githubmofo/TorusGuard/SKILL.md" } }, "instructions": "# TorusGuard Master Security Router\n\n**Principle:** If the browser or client receives it, users can inspect it. Keep secrets, database queries, and authorization strictly on trusted server-side code.\n\n---\n\n## ⚡ Workspace Bootstrap Check\n1. **If `.torusguard/` exists on disk:** Full autonomous governance active. Load `.torusguard/memory/context.json` into prompt context for local-first intelligence. Route commands to specialist workflows below.\n2. **If `.torusguard/` is absent (Autonomous AI Mode):** Operates standalone in any AI agent (Kimi, Antigravity, VS Code, Cursor, Claude Code, Windsurf)", "cost": { "context_tokens": 758 } }

torusguard

About

TorusGuard Master Security Router

⚡ Workspace Bootstrap Check

Specialist Skill Routing (Lazy Loading)

Use it

Manifest

Reviews