safety-guardrails

safety-guardrails is a skill for Claude Code, Codex from 90le/worker-rights-cn. It costs 71 tokens per session (666 once invoked), scanned A, original, Apache-2.0.

A safety guide for handling worker-side labor-rights requests in China. It identifies requests involving fraud, threats, privacy, confidential information, or risky legal tactics.

In plain words
What is it for?
It classifies a request, sets limits on the response, and redirects the work toward lawful evidence preservation, negotiation, agreement review, compensation estimates, or arbitration drafting.
Why use it?
It helps prevent unsafe or unlawful advice, while keeping legitimate help focused on facts, evidence, and appropriate legal review.

Skill for Claude CodeCodex

Written for Claude Code and Codex: shipped in a Claude Code plugin, but also agents/openai.yaml present.

Part of the worker-rights-cn plugin — 11 skills, 5 hooks, 1 MCP server shipped together

Good fit It classifies a request, sets limits on the response, and redirects the work toward lawful evidence preservation, negotiation, agreement review, compensation estimates, or arbitration drafting.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/90le/worker-rights-cn/safety-guardrails
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add 90le/worker-rights-cn --skill safety-guardrails
Clone the repo
git clone --depth 1 https://github.com/90le/worker-rights-cn

Made for: Claude Code, Codex.

Or install worker-rights-cn, the plugin that ships this one along with the rest of its 11 skills, 5 hooks, 1 MCP server.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for safety-guardrails

README.md
[![agentmods](https://agentmods.dev/badge/skills/90le/worker-rights-cn/safety-guardrails/github.svg)](https://agentmods.dev/skills/90le/worker-rights-cn/safety-guardrails)
Your own site
<a href="https://agentmods.dev/skills/90le/worker-rights-cn/safety-guardrails"><img src="https://agentmods.dev/badge/skills/90le/worker-rights-cn/safety-guardrails/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for safety-guardrails

Your own site · 80×15
<a href="https://agentmods.dev/skills/90le/worker-rights-cn/safety-guardrails"><img src="https://agentmods.dev/badge/skills/90le/worker-rights-cn/safety-guardrails.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 71 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 666 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00071 $0.00666
Opus 5 $0.00036 $0.00333
Sonnet 5 $0.00014 $0.00133
Haiku 4.5 $0.00007 $0.00067

Measured 12d ago against content hash 5db79f1b5a25, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

safety-guardrails scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/run_safety_cases.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/worker-rights-cn/skills/safety-guardrails/SKILL.md · 54 lines

How it starts

The opening of the file, as written. The whole thing — 54 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Safety Guardrails

Overview

Classify risky labor-rights requests into refusal, rewrite, lawyer-check, or proceed-with-caution paths. Convert unsafe user intent into lawful evidence-based alternatives with source-card anchors.

Workflow

  1. Read references/redline-policy.json.
  2. Classify the user request into one or more risk_categories.
  3. Apply the strictest decision from the matched categories.
  4. Refuse blocked assistance when required, but offer a lawful alternative.
  5. Route safe follow-up work to evidence-builder, negotiation-coach, agreement-review, arbitration-drafter, or compensation-calculator.
  6. Include source_anchors for legal or procedural bases.

Decision priority:

  • refuse_and_redirect: do not provide instructions, templates, wording, or operational steps for the unsafe act.
  • rewrite_with_limits: remove unsupported facts or risky pressure tactics and draft only a factual, lawful version.
  • lawyer_check: stop short of tactical advice when the action may affect criminal, privacy, confidentiality, non-compete, or evidence-spoliation exposure.
  • proceed_with_caution: continue only with lawful evidence preservation and fact separation.

Output Format

Return:

  • safety_decision: refuse_and_redirect, rewrite_with_limits, lawyer_check, or proceed_with_caution.
  • risk_categories: matched policy category ids and short reasons.
  • blocked_content: what must not be drafted or operationalized.
  • lawful_alternative: concrete safe next step, phrased for the worker.
  • allowed_next_steps: evidence, negotiation, agreement, complaint, arbitration, or lawyer-review actions.
  • source_anchors: SOURCE-ID#artN anchors supporting the boundary or safe path.
  • next_skill: recommended skill after the safety decision.

Drafting Rules

  • Do not give step-by-step help for fabricating, altering, deleting, hiding, or backdating evidence.
  • Do not draft threats, blackmail, harassment, doxxing, public-shaming pressure, or unlawful retaliation.
  • Do not advise copying unrelated company data, source code, customer lists, trade secrets, or third-party personal information.
  • Do not help evade valid confidentiality, service-period, or non-compete duties; instead request scope, payment, release, or lawyer review.
  • Do not promise arbitration, complaint, litigation, or settlement outcomes.
  • Separate confirmed facts from suspected facts, estimates, and claims needing proof.
  • Keep lawful alternatives specific: preserve original records, request written reasons, ask for itemized payments, file a labor-authority complaint, prepare arbitration claims, or seek lawyer review.

Read the full file on GitHub · 54 lines

Files

What ships with it

4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 12d ago First seen · 54 lines · 71 tokens per session scan A 5db79f1b5a25

Subscribe to this mod's changes

safety-guardrails is a skill published in the GitHub repository 90le/worker-rights-cn (94 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 71 tokens to every session and 666 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

compliance

Compliance expert for SOC 2, GDPR, HIPAA, PCI-DSS, and security frameworks.

RightNow-AI/openfang · 22 tokens

dependency-audit

Audits direct and transitive dependencies for license compliance, maintenance health, CVEs, abandoned packages, and bloat. Triggers on: "audit dependencies", "license check", "dependency health", "abandoned packages", "unused dependencies", "license compliance", "supply chain", "dependency risk".

Mathews-Tom/armory · 66 tokens

commercial-legal-pl

Skill do analizy i tworzenia umów według polskiego prawa, ze szczególnym uwzględnieniem umów B2B, IP i IT (body leasing, NDA, wdrożenia, SaaS, przeniesienie praw autorskich, ugody). Powstał w Kancelarii Radców Prawnych Żurawska Piotrowski i Wspólnicy (ktzr.pl). Używaj zawsze gdy użytkownik prosi o przeanalizowanie…

apiotrowski-afk/commercial-legal-pl · 231 tokens

anvil-adsense-audit

A pre-submission audit for websites applying to Google AdSense, Google's advertising programme. It checks policy, content, crawling, privacy, and site-readiness items and records each as Pass, Fail, Unknown, or Not applicable.

PNGTRID/AnvilWiki · 136 tokens

counterparty-guard

A Russian business-counterparty check based on an identification number called an INN. It gathers public records about a company or sole proprietor and presents deal-risk signals.

ilyautov/small-business-ru · 252 tokens

contract-review

A contract-checking tool for Russian law, aimed at small businesses without an in-house lawyer. It reviews supply, service, contractor, NDA, and SaaS agreements and explains risky or missing terms in plain language.

ilyautov/small-business-ru · 175 tokens