OpenClaw Governance Skill Suite

English | 简体中文

A GitHub-ready suite of four OpenClaw skills for turning ad hoc agent usage into a governed, reusable system.

Quick Start

Option 1 — Copy one skill into your OpenClaw workspace

Copy the skill folder you want into:

~/.openclaw/workspace/skills/

Example:

cp -R skills/context-manifest ~/.openclaw/workspace/skills/

Option 2 — Use this repository as a maintained suite

Keep this repository as a source-of-truth collection, then copy or sync specific skills into your local OpenClaw workspace when needed.

Included skills

context-manifest

Govern project context with MANIFEST.md.

Key capabilities:

• define project truth with Canonical / Domain Map / Archive / Ignore / Sensitive
• route tasks into the minimum useful context
• support scoped task packages for sub-agents
• standardize project-level Operating Rules
• apply plan / approve / execute and uncertainty handling patterns

trust-boundary

Isolate untrusted input and resist prompt injection.

Key capabilities:

• classify trusted / semi-trusted / untrusted inputs
• handle webpages, emails, PDFs, OCR, external repos, and logs safely
• prevent untrusted content from authorizing high-impact actions
• keep evidence separate from approval, trust, and project truth

scheduled-workflows

Design artifact-first recurring automations.

Key capabilities:

• model durable scheduled workflows with clear trigger / input / process / output / delivery / failure boundaries
• standardize artifact naming and retention
• support notify-on-change, notify-on-failure, and low-noise delivery
• handle retry, degradation, and partial-success states explicitly

skill-orchestrator

Route, compose, extend, and fall back across skills.

Key capabilities:

• prefer reuse -> combine -> extend -> create
• design explicit skill stacks for complex tasks
• decide when to extend an existing skill versus creating a new one
• handle stack conflicts and graceful fallback paths

Suggested adoption order

1. Start with context-manifest
2. Add trust-boundary when external inputs are involved
3. Use scheduled-workflows for recurring automations
4. Use skill-orchestrator when work spans multiple skills or new-skill decisions

Onboarding and operational docs

Start here if this is your first time adopting the suite:

• Getting Started
• Skill Examples
• Anti-Patterns
• Cross-Skill Playbook
• First Adoption Checklist

Reusable starter files:

• MANIFEST template
• Scheduled workflow template

Installation / Use

This repository is a skill suite repository, not a single packaged .skill.

Typical usage flow:

1. clone or download this repository
2. choose the skill you need from skills/
3. copy that skill folder into your OpenClaw workspace
4. iterate locally as needed

Example trigger patterns:

• “Use trust-boundary before handling external PDFs or webpages.”
• “Use context-manifest to define project truth and reduce context bloat.”
• “Use skill-orchestrator before creating a new skill from scratch.”
• “Use scheduled-workflows to design a recurring artifact-first automation.”

Repository layout

openclaw-governance-skill-suite/
├── LICENSE
├── README.md
├── README.zh-CN.md
├── ROADMAP.md
├── .gitignore
└── skills/
    ├── context-manifest/
    ├── trust-boundary/
    ├── scheduled-workflows/
    └── skill-orchestrator/

Status

These skills are currently validated draft skills.

They were iteratively refined for:

• clearer scope boundaries
• reusable references
• safer execution patterns
• lower-noise automation design

They are ready for real-world trial use and further iteration.

Important note

This repository is an exported copy for GitHub publication.

It does not replace or modify the original local skills in:

~/.openclaw/workspace/skills/

Each skill remains self-contained, with its own SKILL.md and references/ directory.

Release

• Latest release: https://github.com/ziiinian/openclaw-governance-skill-suite/releases/latest
• First public release: https://github.com/ziiinian/openclaw-governance-skill-suite/releases/tag/v1.0.0

License

This repository currently uses the MIT License for easy public sharing. Change it before publishing if you want a different licensing model.