How do I install the Yana AI skill?

Run npx skillmds@latest add yanacuti1121/yana-ai-2 in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Yana AI safe to use?

This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: WARNING, Skill Scanner: WARNING. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Yana AI?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Yana AI free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under Apache-2.

Who published Yana AI?

yanacuti1121 (@yanacuti1121) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Yana AI by yanacuti1121 — Agent Skill

Name: Yana AI
Author: yanacuti1121

Yana AI

Sovereign-grade safety OS for AI coding agents. 66 hooks, 2,025 skills, L1 memory, circuit breakers, and cross-engine enforcement — blocks rm -rf, force push, pipe-to-shell, and 40+ attack vectors before they reach your repo.

yanacuti1121 Updated Sep 22, 2026 2 repo stars

File contents

Yana AI — Sovereign Safety OS

Yana AI is an agent operating system layered on top of Claude Code (and other AI coding agents). It intercepts dangerous commands, enforces safety rules, manages persistent memory, and provides 2,025+ reusable skills.

What Yana AI does

When loaded, Yana AI automatically:

Blocks 40+ attack vectors via 60 pre/post hooks — rm -rf, force push, pipe-to-shell, path traversal, supply chain attacks, and more
Enforces rules across all agent actions using a 9-layer middleware pipeline (injection scan → blast radius → permission check → egress → sanitize → PII scrub → size cap → audit log)
Persists memory in a 2-tier system: L1 Atomic (permanent, hash-chained) and L2 session facts
Provides 2,025 skills covering frontend, backend, IaC, AI/LLM, security, multi-agent, K8s, WebAssembly, and more

Install

# Install via skills CLI
npx skills add yanacuti1121/yana-ai

# Or install directly into .claude/
curl -L https://github.com/yanacuti1121/yana-ai/releases/latest/download/yana-ai-latest.zip -o yana-ai.zip
unzip yana-ai.zip -d .claude/

Quick commands

Command	What it does
`/status`	Show hooks, memory, and gate status
`/audit`	Run full security audit on repo
`/quick-commit`	Safe commit with gate checks
`/session-wrap`	Persist session state to L1 memory
`/smart-fix`	Auto-fix with feedback loop
`/code-review`	Multi-agent code review

Safety hooks overview

Yana AI hooks fire on every tool call:

PreToolUse:
  - safe-run.sh         — blocks 40+ dangerous command patterns
  - scope-guard.sh      — enforces declared file scope
  - risk-scorer.sh      — rates action risk 0–100
  - token-budget-guard  — prevents runaway token usage

PostToolUse:
  - session-checkpoint  — saves state every 5 tool calls
  - audit-logger        — appends to Merkle hash-chain log

Stop:
  - truth-gate.sh       — warns on completion claims without evidence

Memory system

# Save a fact to L1 (persists across sessions)
bash core/scripts/add-fact.sh "tag" "fact content" "high"

# Search L1 memory
bash core/scripts/search-facts.sh "tag"

# View session memory
bash core/scripts/list-session-facts.sh

Hard-blocked commands

Yana AI will always block these regardless of context:

rm -rf          git push --force     eval "$user_input"
curl ... | bash  DROP TABLE           chmod 777 core/

Cross-engine enforcement

Yana AI enforces safety on all AI coding agents via a shared safe-run.sh proxy:

Claude Code — native hooks
Codex — CODEX_SHELL_EXEC_HOOK → safe-run.sh
Cursor — .cursor/rules/ enforcement
Aider — --before-exec hook

---name: yana-ai-2description: Sovereign-grade safety OS for AI coding agents. 66 hooks, 2,025 skills, L1 memory, circuit breakers, and cross-engine enforcement — blocks rm -rf, force push, pipe-to-shell, and 40+ attack vectors before they reach your repo.license: Apache-2.0---# Yana AI — Sovereign Safety OSYana AI is an agent operating system layered on top of Claude Code (and other AI coding agents). It intercepts dangerous commands, enforces safety rules, manages persistent memory, and provides 2,025+ reusable skills.## What Yana AI doesWhen loaded, Yana AI automatically:- **Blocks 40+ attack vectors** via 60 pre/post hooks — `rm -rf`, force push, pipe-to-shell, path traversal, supply chain attacks, and more- **Enforces rules** across all agent actions using a 9-layer middleware pipeline (injection scan → blast radius → permission check → egress → sanitize → PII scrub → size cap → audit log)- **Persists memory** in a 2-tier system: L1 Atomic (permanent, hash-chained) and L2 session facts- **Provides 2,025 skills** covering frontend, backend, IaC, AI/LLM, security, multi-agent, K8s, WebAssembly, and more## Install```bash# Install via skills CLInpx skills add yanacuti1121/yana-ai# Or install directly into .claude/curl -L https://github.com/yanacuti1121/yana-ai/releases/latest/download/yana-ai-latest.zip -o yana-ai.zipunzip yana-ai.zip -d .claude/```## Quick commands| Command | What it does ||---------|-------------|| `/status` | Show hooks, memory, and gate status || `/audit` | Run full security audit on repo || `/quick-commit` | Safe commit with gate checks || `/session-wrap` | Persist session state to L1 memory || `/smart-fix` | Auto-fix with feedback loop || `/code-review` | Multi-agent code review |## Safety hooks overviewYana AI hooks fire on every tool call:```PreToolUse: - safe-run.sh — blocks 40+ dangerous command patterns - scope-guard.sh — enforces declared file scope - risk-scorer.sh — rates action risk 0–100 - token-budget-guard — prevents runaway token usagePostToolUse: - session-checkpoint — saves state every 5 tool calls - audit-logger — appends to Merkle hash-chain logStop: - truth-gate.sh — warns on completion claims without evidence```## Memory system```bash# Save a fact to L1 (persists across sessions)bash core/scripts/add-fact.sh "tag" "fact content" "high"# Search L1 memorybash core/scripts/search-facts.sh "tag"# View session memorybash core/scripts/list-session-facts.sh```## Hard-blocked commandsYana AI will always block these regardless of context:```rm -rf git push --force eval "$user_input"curl ... | bash DROP TABLE chmod 777 core/```## Cross-engine enforcementYana AI enforces safety on all AI coding agents via a shared `safe-run.sh` proxy:- **Claude Code** — native hooks- **Codex** — `CODEX_SHELL_EXEC_HOOK` → safe-run.sh- **Cursor** — `.cursor/rules/` enforcement- **Aider** — `--before-exec` hook

Frequently asked questions