← 返回 Skills 市场

AI Reliability Layer

Name: AI Reliability Layer
Author: renoblabs

作者 renoblabs · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install ai-reliability-layer

功能描述

Fix broken LLM output, validate AI responses, generate guaranteed structured JSON. Three micro-services for making AI output reliable. Use when LLM output is...

使用说明 (SKILL.md)

AI Reliability Layer

Three pay-per-call micro-services that fix the most common AI output problems.

Services

/fix-json — Fix Broken LLM JSON

Handles: trailing commas, single quotes, truncated output, markdown code fences, missing brackets.

POST /x402s/fix-json
Body: {"input": "{\"name\": \"test\",}", "schema": {...optional...}}
Response: {"output": {"name": "test"}, "method": "trailing_comma", "fixed": true}
Price: $0.001 USDC

/validate-output — Enforce Rules on AI Output

Check length, format, schema compliance, blocked patterns.

POST /x402s/validate-output
Body: {"output": "...", "rules": {"max_length": 500, "format": "json", "schema": {...}}}
Response: {"valid": true/false, "violations": [...]}
Price: $0.001 USDC

/generate-structured — Guaranteed Typed Response

Prompt + JSON schema in, valid typed JSON out. Uses LLM with auto-fix.

POST /x402s/generate-structured
Body: {"prompt": "List 3 fruits with colors", "schema": {"type": "array", "items": {"type": "object", "properties": {"name": {}, "color": {}}}}}
Response: {"output": [...], "model": "qwen3.5:9b", "valid": true}
Price: $0.01 USDC

Payment

All services use x402 protocol — USDC on Base. No API keys, no subscriptions.

Part of AEA Arena

Built by the Reno Labs agent fleet. More services at /x402s/catalog.

安全使用建议

This skill describes external paid microservices but leaves out critical details. Before installing or using it: (1) ask the publisher for the full base URL(s) for /x402s endpoints and documentation of the x402 payment flow; (2) confirm how payments are made and whether a wallet private key or RPC credentials are required — do not provide private keys unless you understand and trust the operator; (3) avoid sending sensitive or private user data to the service until you verify its privacy/security practices; (4) request an explanation of why no environment credentials are declared despite on-chain billing; (5) prefer local or well-known libraries to fix/validate JSON if you cannot confirm the service's provenance; (6) if you allow autonomous agent use, restrict or audit any automatic calls that could incur on-chain payments. If the publisher cannot provide clear answers and a verifiable service domain, treat the skill as untrusted.

功能分析

Type: OpenClaw Skill Name: ai-reliability-layer Version: 1.0.0 The bundle consists solely of documentation (SKILL.md) and metadata (_meta.json) describing three AI utility micro-services for JSON fixing and validation. There is no executable code, and the instructions are limited to describing API endpoints and a crypto-based payment protocol (x402 on Base); no malicious intent, data exfiltration, or prompt injection risks were identified.

能力标签

cryptocan-make-purchases

能力评估

⚠ Purpose & Capability

The claimed purpose (fixing/validating/generating structured JSON) matches the described API endpoints, but the SKILL.md advertises on-chain payment (x402 / USDC on Base) while the metadata requests no credentials or wallet access. A pay-per-call external service normally requires a full base URL and signing/payment credentials; the absence of those is incongruent.

⚠ Instruction Scope

SKILL.md lists example POST bodies and responses but provides only relative paths (/x402s/...) with no host, no HTTP headers, and no instructions on how to perform x402 payments or sign transactions. It also gives no guidance about redacting or protecting sensitive user data before sending it to an external service. The instructions are therefore incomplete and leave broad, unspecified behavior to the agent.

✓ Install Mechanism

This is an instruction-only skill with no install spec and no code files. That's low-risk from an install standpoint: nothing is written to disk or fetched automatically.

⚠ Credentials

The skill requests no environment variables or credentials, yet describes a paid API on a blockchain protocol. To actually pay for calls the agent (or user) would need a wallet/private key, RPC access, or some payment mechanism — none of which are declared. This mismatch is disproportionate and unexplained. Also there is potential for data exfiltration because network calls are implied but not constrained.

✓ Persistence & Privilege

The skill is not always-on and defaults to user invocation; it does not request elevated persistence or system-wide config changes. No privilege escalation is evident from the metadata.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install ai-reliability-layer
安装完成后，直接呼叫该 Skill 的名称或使用 /ai-reliability-layer 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Initial release of AI Reliability Layer micro-services - Three pay-per-call endpoints: /fix-json, /validate-output, /generate-structured - /fix-json: Automatically corrects malformed JSON from LLMs (e.g., trailing commas, missing brackets) - /validate-output: Checks AI output for length, format, schema compliance, and blocked patterns - /generate-structured: Produces valid, typed JSON responses from prompts using any custom schema - All services require no API keys or subscriptions; payments handled via the x402 protocol (USDC on Base) - Part of Reno Labs’ AEA Arena service fleet

元数据

Slug ai-reliability-layer

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题