governance-wrapper
由 mikeholownych 提供
Coding Agents & IDEs简介
Mandatory strict-mode execution wrapper for autonomous operations with evidence logging and policy enforcement.
描述
该技能的功能
Wraps autonomous agent operations in a strict execution environment, enforcing model lock, network and skill allowlists, and concurrent subagent limits. Every execution emits a hash-chained evidence record for auditing. Policy violations are blocked and recorded automatically.
Hash-chained append-only evidence logging makes execution history tamper-evident in a way standard logging cannot guarantee.
何时使用
- Auditing autonomous agent actions in compliance-sensitive pipelines
- Preventing unauthorized model or skill substitutions during automation runs
- Restricting outbound network access from agent workflows to an approved list
- Capping concurrent subagent spawns to control resource consumption
- Generating tamper-evident logs of agent executions for post-incident review
示例工作流
User asks: Run the summarization skill on this document with full audit trail
- 1
Invoke governance_wrapper.py with --requested-skill, --system-prompt, and --input-context parameters
- 2
Wrapper validates the requested skill against the tool surface manifest allowlist
- 3
Wrapper locks execution to opencode/big-pickle at temperature 0.0 with no fallback
- 4
Wrapper enforces network allowlist for any outbound HTTP calls made during execution
- 5
Emits execution-evidence.v1 record and appends it to the hash-chained evidence log
Skill executes within policy bounds; evidence record appended to the audit log with hash chain intact
安装
clawhub install mikeholownych/governance-wrapper运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。