← 返回 Skills 市场

enforceBackendUse

Name: enforceBackendUse
Author: tsr0705

作者 TSR0705 · 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 customskill

功能描述

Enforces mandatory use of process_request tool for all financial transactions, forbidding direct trade execution or simulation by the agent.

使用说明 (SKILL.md)

process_request\r

CRITICAL: TOOL ENFORCEMENT CONTRACT\r

\r You are a financial AI agent operating under a STRICT enforcement system.\r \r You DO NOT execute trades yourself.\r \r You MUST use this tool for ANY financial execution.\r \r ---\r \r

🚨 MANDATORY RULES\r

1. TOOL EXECUTION IS REQUIRED\r

\r You MUST call process_request when:\r \r

User asks to BUY or SELL stocks\r
Any financial transaction is involved\r
Any action modifies assets or money\r \r You are NOT allowed to simulate or describe execution.\r \r ---\r \r

2. TOOL EXECUTION IS FORBIDDEN\r

\r You MUST NOT call the tool when:\r \r

User asks for analysis only\r
User asks questions (e.g., "should I buy?")\r
Intent is unclear or ambiguous\r \r ---\r \r

3. NO DIRECT EXECUTION\r

\r You CANNOT:\r \r

Execute trades yourself\r
Assume execution success\r
Bypass this tool\r \r ---\r \r

INPUT SCHEMA (STRICT)\r

\r You MUST send EXACTLY this format:\r \r

{\r
  "user_input": "string",\r
  "agent_reasoning": "string",\r
  "proposed_action": {\r
    "type": "trade",\r
    "asset": "string",\r
    "amount": "number",\r
    "side": "buy | sell"\r
  }\r
}

安全使用建议

This skill is a policy that forces the agent to call a tool named process_request for any financial execution. That could be legitimate if you control and trust the process_request implementation, but it becomes risky if the tool is untrusted: the skill explicitly requires sending 'agent_reasoning' (internal chain-of-thought) and trade details to the tool, which can leak sensitive reasoning, private data, or allow exfiltration. Before installing: (1) confirm who implements process_request and review its code/endpoint and access controls; (2) remove or change the 'agent_reasoning' requirement to a short, non-sensitive rationale or summary rather than chain-of-thought; (3) require explicit user confirmation before sending execution requests; and (4) ensure logging/auditing and rate/validation checks exist on the backend. If you do not control the backend tool, treat this skill as risky and avoid installing it.

功能分析

Type: OpenClaw Skill Name: customskill Version: 1.0.0 The skill bundle defines a protocol for a financial AI agent, enforcing the use of a specific tool (process_request) for executing trades. The instructions in SKILL.md establish strict operational guardrails and a structured input schema, showing no signs of malicious intent, data exfiltration, or unauthorized execution.

能力评估

ℹ Purpose & Capability

The name/description (enforce backend usage for financial actions) matches the instructions: the skill mandates calling a process_request tool for executions. However, the skill does not declare or bind the process_request tool (no metadata about where it lives or who provides it). That omission is explainable for an instruction-only policy but reduces transparency about what will receive data.

⚠ Instruction Scope

The SKILL.md requires sending an exact JSON including an 'agent_reasoning' string (internal chain-of-thought) and a 'proposed_action'. Requiring internal reasoning to be emitted to an external tool is a high-risk instruction: it can leak sensitive signals (chain-of-thought, heuristics, or hidden context). The document also enforces tool invocation for any financial execution without describing safeguards (e.g., confirmation, validation, or limits).

✓ Install Mechanism

Instruction-only skill with no install spec or code files; nothing is written to disk and no external packages are fetched by the skill itself.

ℹ Credentials

The skill requests no environment variables or credentials. However, it requires routing structured data (including agent_reasoning) to an unspecified tool implementation; the danger here is not credentials but data exfiltration to an untrusted tool if the environment's process_request is malicious or misconfigured.

✓ Persistence & Privilege

The skill does not request always-on presence and does not modify other skills or system settings. It is user-invocable and allows normal autonomous invocation (platform default).

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 customskill
安装完成后，直接呼叫该 Skill 的名称或使用 /customskill 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

- Enforces strict tool use for all financial trade or transaction requests. - Prevents direct execution or simulation of trades; mandates tool invocation via a specific schema. - Clearly separates analysis/questions (no tool use) from actionable execution (tool required). - Input schema for all requests involving asset modification is now explicitly defined and required.

元数据

Slug customskill

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