Releases: desek/governance
governance: v1.3.0
checkpoint-read: v1.0.0
governance: v1.2.0
🚀 Automating Governance: The Rise of /checkpoint-commit
Keeping track of architectural decisions and work-in-progress state just got a whole lot smoother. In this update, we’ve overhauled how the Governance Skill handles iterative work by introducing a dedicated slash command, automating PR title validation to keep our history clean, and refining our document creation guidelines to save you (and your AI agents) unnecessary build time. ✨
🤖 New Skill: /checkpoint-commit
To solve the friction of manual work preservation, we've introduced the /checkpoint-commit skill. This command automates the entire process of analyzing changes, validating .gitignore to prevent repository bloat, and creating a formatted commit linked to your active Change Request (CR). No more manual multi-step workflows—just run the command and let the agent handle the "Golden Summary" and safety checks.
The command is implemented as a standalone skill, ensuring separation of concerns from the core governance documentation tools. It even features context-awareness to auto-detect your CR ID from branch names or recent file history.
# Simply run the command to preserve your work
/checkpoint-commit CR-0010 "Refactor skill logic"- Source:
skills/checkpoint-commit/SKILL.md(lines 1-99),skills/checkpoint-commit/version.txt(lines 1-2)
🛡️ Automated PR Title Validation
Maintaining a linear, readable history is critical for automated release tools like Release Please. We've added a GitHub Actions workflow that strictly enforces Conventional Commits for all Pull Request titles. This acts as a quality gate, blocking merges that don't follow the <type>[scope]: <description> format, ensuring that our changelogs and version bumps are always accurate. 🚀
# Examples of valid titles now required:
feat(auth): add login support
fix: resolve memory leak in worker
docs(governance): update checkpoint guide- Source:
.github/workflows/pr-title.yml(lines 1-48),docs/cr/CR-0008-pr-title-validation.md(lines 1-343)
🔗 Seamless Checkpoint Hooks
We've integrated the checkpoint workflow directly into the Claude Code lifecycle using prompt-based hooks. By configuring a Stop hook, the agent can now evaluate if there are uncommitted changes before it finishes a session. If changes exist, it triggers a "Seamless Handover," suggesting or executing /checkpoint-commit automatically to ensure no work is ever lost.
- Source:
skills/governance/reference/checkpoint-hooks.md(lines 1-120),README.md(lines 29-53)
📖 Refined Governance Guidelines
To optimize agent performance, we've updated the Governance Skill instructions to explicitly label CR and ADR creation as documentation-only tasks. This prevents agents from wasting time running build pipelines or test suites when they are simply writing Markdown. Additionally, we've clarified that governance documents should omit metadata fields like copyright and version to keep the frontmatter clean and focused.
- Source:
skills/governance/SKILL.md(lines 20-100),AGENTS.md(lines 8-120)
Minor Updates & Housekeeping
- Global Config: Added
vercel-labs/skillsto.deepwikiand updated.gitignoreto exclude.envfiles.- Source:
.deepwiki(line 5),.gitignore(lines 2-10).
- Source:
- Dependency Management: Updated
release-please-config.jsonto support multi-package releases with separate pull requests for each skill.- Source:
release-please-config.json(lines 1-18).
- Source:
- Versioning: Bumped
governanceskill to1.2.0and initializedcheckpoint-commitat0.0.0.- Source:
.release-please-manifest.json(lines 1-4).
- Source:
Meta Description: Streamline your K8s governance with the new /checkpoint-commit slash command, automated PR title validation, and optimized Claude Code hook integrations.
Primary Category: Developer Tools
Tags:
- By Project: Governance Skill, Checkpoint-Commit
- By Tool/Technology: Claude Code, GitHub Actions, Release Please, Git
- By Concept: Infrastructure as Code (IaC), Automation, Conventional Commits, Change Management
checkpoint-commit: v1.0.0
🚀 Automating Governance: The Rise of /checkpoint-commit
Keeping track of architectural decisions and work-in-progress state just got a whole lot smoother. In this update, we’ve overhauled how the Governance Skill handles iterative work by introducing a dedicated slash command, automating PR title validation to keep our history clean, and refining our document creation guidelines to save you (and your AI agents) unnecessary build time. ✨
🤖 New Skill: /checkpoint-commit
To solve the friction of manual work preservation, we've introduced the /checkpoint-commit skill. This command automates the entire process of analyzing changes, validating .gitignore to prevent repository bloat, and creating a formatted commit linked to your active Change Request (CR). No more manual multi-step workflows—just run the command and let the agent handle the "Golden Summary" and safety checks.
The command is implemented as a standalone skill, ensuring separation of concerns from the core governance documentation tools. It even features context-awareness to auto-detect your CR ID from branch names or recent file history.
# Simply run the command to preserve your work
/checkpoint-commit CR-0010 "Refactor skill logic"- Source:
skills/checkpoint-commit/SKILL.md(lines 1-99),skills/checkpoint-commit/version.txt(lines 1-2)
🛡️ Automated PR Title Validation
Maintaining a linear, readable history is critical for automated release tools like Release Please. We've added a GitHub Actions workflow that strictly enforces Conventional Commits for all Pull Request titles. This acts as a quality gate, blocking merges that don't follow the <type>[scope]: <description> format, ensuring that our changelogs and version bumps are always accurate. 🚀
# Examples of valid titles now required:
feat(auth): add login support
fix: resolve memory leak in worker
docs(governance): update checkpoint guide- Source:
.github/workflows/pr-title.yml(lines 1-48),docs/cr/CR-0008-pr-title-validation.md(lines 1-343)
🔗 Seamless Checkpoint Hooks
We've integrated the checkpoint workflow directly into the Claude Code lifecycle using prompt-based hooks. By configuring a Stop hook, the agent can now evaluate if there are uncommitted changes before it finishes a session. If changes exist, it triggers a "Seamless Handover," suggesting or executing /checkpoint-commit automatically to ensure no work is ever lost.
- Source:
skills/governance/reference/checkpoint-hooks.md(lines 1-120),README.md(lines 29-53)
📖 Refined Governance Guidelines
To optimize agent performance, we've updated the Governance Skill instructions to explicitly label CR and ADR creation as documentation-only tasks. This prevents agents from wasting time running build pipelines or test suites when they are simply writing Markdown. Additionally, we've clarified that governance documents should omit metadata fields like copyright and version to keep the frontmatter clean and focused.
- Source:
skills/governance/SKILL.md(lines 20-100),AGENTS.md(lines 8-120)
Minor Updates & Housekeeping
- Global Config: Added
vercel-labs/skillsto.deepwikiand updated.gitignoreto exclude.envfiles.- Source:
.deepwiki(line 5),.gitignore(lines 2-10).
- Source:
- Dependency Management: Updated
release-please-config.jsonto support multi-package releases with separate pull requests for each skill.- Source:
release-please-config.json(lines 1-18).
- Source:
- Versioning: Bumped
governanceskill to1.2.0and initializedcheckpoint-commitat0.0.0.- Source:
.release-please-manifest.json(lines 1-4).
- Source:
Meta Description: Streamline your K8s governance with the new /checkpoint-commit slash command, automated PR title validation, and optimized Claude Code hook integrations.
Primary Category: Developer Tools
Tags:
- By Project: Governance Skill, Checkpoint-Commit
- By Tool/Technology: Claude Code, GitHub Actions, Release Please, Git
- By Concept: Infrastructure as Code (IaC), Automation, Conventional Commits, Change Management
governance: v1.1.0
Governance v1.1: Smarter Agent Memory & Standardized Compliance
We are excited to announce the release of Governance v1.1.0! This update is all about maturing the workflow for both AI agents and human contributors. We've introduced a robust "short-term memory" system for development via iterative checkpoints, standardized file metadata across the entire repository for better compliance, and improved discoverability for LLMs.
💾 Iterative Checkpoints: Short-Term Memory for Agents
Developing complex features often requires saving state without polluting the permanent project history. To solve the issue of messy work-in-progress commits, we've introduced a formal Checkpoint Instruction. This gives agents and developers a structured way to capture "short-term memory" on active branches—preserving staged, unstaged, and untracked changes—while ensuring the final merge ("long-term memory") remains clean and semantic.
The new workflow forces a hygiene check before every save. It requires analyzing changes and explicitly updating .gitignore to prevent temporary build artifacts from creeping into the repository. Once the work is ready to merge, these checkpoints are designed to be semantically compressed into a single, coherent squash merge.
The new Checkpoint Workflow:
# 1. Analyze state
git diff --staged
git diff
git ls-files --others --exclude-standard
# 2. Update .gitignore (Mandatory step!)
# 3. Create the checkpoint
git add -A
git commit -m "checkpoint(CR-xxxx): {summary}"- Source:
skills/governance/SKILL.md(lines 80-101) andskills/governance/reference/checkpoint.md(lines 1-111)
⚖️ Standardized Copyright & Metadata Headers
As the repository grows, maintaining clear attribution and licensing becomes critical. We have implemented a rigorous standardization policy requiring copyright headers on all files. This isn't just about legal protection; it's about consistency. We've moved away from ad-hoc comments to structured YAML frontmatter for Markdown files, ensuring that file descriptions, versioning, and copyright ownership are machine-readable and uniform.
For every file type, there is now a strict format defined in AGENTS.md. Markdown files, in particular, now carry rich metadata that describes the file's intent, making it easier for both humans and agents to understand the context of a document before reading its content.
New Markdown Frontmatter Standard:
---
name: file-identifier
description: Brief description of the file's purpose
metadata:
copyright: Copyright Daniel Grenemark 2026
version: "0.0.1"
---New TOML/Text Standard:
# Copyright Daniel Grenemark 2026- Source:
AGENTS.md(lines 69-95) anddocs/cr/CR-0005-copyright-headers.md(lines 1-464)
🤖 LLM Discoverability with llms.txt
To make our documentation more accessible to AI models and automated tools, we have added an llms.txt file. This acts as a manifest for the repository, providing a clean, curated list of high-value documentation—specifically the Change Requests (CRs). This allows LLMs to quickly index the history of architectural decisions without traversing the entire file tree.
# Governance Documentation
> Documentation for the governance repository...
## Change Requests
- [CR-0001: Refactor Governance Skill](cr/CR-0001-refactor-governance-skill.md)
...
- Source:
docs/llms.txt(lines 1-12)
Minor Updates & Housekeeping
- Tooling Updates: The
junie-clitool has been updated from624.1.0to704.1.0in themise.tomlconfiguration.- Source:
mise.toml(line 3)
- Source:
- Release Configuration: We've cleaned up the
release-please-config.jsonto streamline version tagging, removing the hardcodedrelease-asoverride now that 1.0.0 is passed.- Source:
release-please-config.json(line 8)
- Source:
- Changelog: The
CHANGELOG.mdhas been updated to reflect the v1.1.0 release and the inclusion of copyright headers.- Source:
skills/governance/CHANGELOG.md(lines 11-16)
- Source:
Discovery
Meta Description:
Governance v1.1.0 introduces iterative Git checkpoints for agent memory, standardized copyright metadata for all files, and an llms.txt manifest for better AI discoverability.
Primary Category:
DevOps & Engineering Governance
Tags:
- Project: Governance, Governance Skill
- Tool/Technology: Git, Release Please, Mise, Markdown
- Concept: Architecture Decision Records (ADR), Change Requests (CR), Developer Workflow, Compliance, AI Agents
governance: v1.0.0
🎉 governance v1.0.0 — Initial Release
We're excited to announce the first release of governance — reusable agent skills for AI-assisted project governance!
✨ What's Included
Governance Skill
A comprehensive skill that enables AI coding agents to create and manage governance documentation:
Architecture Decision Records (ADRs)
- MADR-based template for documenting architectural decisions
- Structured sections: Context, Decision Drivers, Options, Considered Options
- Clear lifecycle management: proposed → accepted → deprecated/superseded
- Reference guide with best practices
Change Requests (CRs)
- Comprehensive 500+ line template for requirement changes
- RFC 2119 keywords (MUST, SHOULD, MAY) for unambiguous requirements
- Gherkin-format acceptance criteria (Given-When-Then)
- Mermaid diagram support for visualizations
- Built-in test strategy section
- Quality standards compliance checklist
- Scope boundaries with "here, but not further" principle
🚀 Quick Start
Install using the skills CLI:
# Install all skills from this repository
npx skills add desek/governance
# Install to specific agents
npx skills add desek/governance -a claude-code -a opencode
# Install globally
npx skills add desek/governance -g🤖 Supported Agents
Works with many AI coding agents including:
- Claude Code
- Codex
- Cursor
- OpenCode
- GitHub Copilot
- And more!
📚 Documentation
- README — Installation and usage
- CONTRIBUTING — How to contribute
- Skill Documentation — Detailed skill guide
📄 License
Apache License, Version 2.0
AI Disclaimer: This project uses AI and may produce inaccurate results. Verify all critical outputs.
Full Changelog: https://github.com/desek/governance/commits/governance-v1.0.0