skills/update-changelog/SKILL.md
Expert skill for maintaining a Keep a Changelog formatted CHANGELOG.md file. Use this skill whenever you add features, fix bugs, or release a new version. You MUST use this skill to record any changes that have a user-facing impact. It handles categorization (Added, Changed, Fixed, etc.), semantic versioning, and reverse-chronological ordering with surgical precision.
npx skillsauth add harshitsinghbhandari/domain-expansion update-changelogInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Maintain accurate, user-facing changelog entries as you build. This skill helps you document new features, fixes, enhancements, and breaking changes in a structured format that follows Keep a Changelog.
The changelog is your project's narrative—it tells users what changed and why. AI agents should document their work just as humans do.
Use this skill whenever you:
The CHANGELOG.md file has this structure:
# Changelog
## [Unreleased]
### Added
### Changed
### Deprecated
### Removed
### Fixed
### Security
## [Version] - [Date]
### Added
...
Important sections:
[Unreleased] — changes that haven't been released yet (always at the top)Decide which category your change belongs to:
| Category | When to Use | Example | |----------|-----------|---------| | Added | New feature or capability | "Add priority-based task queuing for agents" | | Changed | Behavior or API modification | "Improve agent coordination algorithm" | | Deprecated | Mark something for future removal | "Deprecate legacy executor API" | | Removed | Delete functionality | "Remove deprecated tmux executor support" | | Fixed | Bug resolution | "Fix agent crash when PR has no comments" | | Security | Vulnerability or safety fix | "Fix privilege escalation in plugin system" |
Write entries from the user's perspective, not implementation details.
✅ Good entries:
❌ Avoid:
If relevant, link to the PR, issue, or commit:
- Added new plugin system for extensible agent behaviors ([PR #42](https://github.com/your-org/your-repo/pull/42))
- Fixed memory leak in event listener ([Issue #89](https://github.com/your-org/your-repo/issues/89))
Find the ## [Unreleased] section at the top of the file.
Add your entry as a bullet point under the relevant category in [Unreleased]:
## [Unreleased]
### Added
- New feature description goes here
- Another feature
### Fixed
- Bug fix description goes here
docs: update changelog for [feature name]What you did: Built a new plugin system for agents.
The entry you add to [Unreleased]/Added:
- Add plugin system with extensible hooks for custom agent behaviors ([PR #42](https://github.com/your-org/your-repo/pull/42))
In context (CHANGELOG.md):
## [Unreleased]
### Added
- Add plugin system with extensible hooks for custom agent behaviors ([PR #42](https://github.com/your-org/your-repo/pull/42))
- Improve agent task decomposition for better parallelization
### Fixed
- Fix agents hanging on repos with very large file trees
What you did: Resolved agents crashing when processing empty PRs.
The entry you add to [Unreleased]/Fixed:
- Fix agent crash when processing PRs with no body or comments
When you're ready to formalize changes as a release:
Use Semantic Versioning:
Replace [Unreleased] with a new version heading:
## [0.2.0] - 2025-03-26
### Added
- Add plugin system with extensible hooks for custom agent behaviors
- Improve agent task decomposition for better parallelization
### Fixed
- Fix agent crash when processing PRs with no body or comments
## [Unreleased]
### Added
### Changed
### Deprecated
### Removed
### Fixed
### Security
If you have a version summary table, update it:
| Version | Date | Highlights |
|---------|------|-----------|
| 0.2.0 | 2025-03-26 | Plugin system + better decomposition |
| 0.1.0 | 2025-03-01 | Foundation + initial setup |
git add CHANGELOG.md
git commit -m "release: v0.2.0"
git tag v0.2.0
git push origin main --tags
| Mistake | Wrong | Right | |---------|-------|--------| | Implementation details | "Refactor executor class" | "Improve executor performance by 30%" | | Missing category | Entry floating in the file | Clearly under Added/Fixed/Changed/etc. | | Too vague | "Various improvements" | "Improve agent coordination during peak load" | | No user impact | "Update dependencies" | Only if users are affected (e.g., "Drop Python 3.8 support") | | Typos | "Imprve agent responsivness" | Spell-check before committing |
You just added a feature: agents can now run custom pre-flight checks before starting work.
Your changelog entry:
### Added
- Allow agents to run custom pre-flight checks before task execution
In context:
## [Unreleased]
### Added
- Allow agents to run custom pre-flight checks before task execution
✅ Why this works:
When an AI agent is updating the changelog:
CHANGELOG.md (root directory).claude/skills/update-changelog/SKILL.md (or your skill directory)# Workflow
1. Make changes to your codebase
2. Run this skill to add entries to CHANGELOG.md
3. Review the entries for clarity
4. Commit: git add CHANGELOG.md && git commit -m "docs: update changelog"
5. When releasing: create version section, commit, and tag
# Categories (in order)
- Added (new features)
- Changed (behavior changes)
- Deprecated (warn about removal)
- Removed (deleted features)
- Fixed (bug fixes)
- Security (vulnerabilities)
# Entry format
- Short, user-facing description
- Optional: link to PR/issue
- Start with action verb (Add, Improve, Fix, Allow, etc.)
Last updated: March 2025
Maintained with: Agent Orchestrator by Composio (MIT License)
development
Aggressive user-flow and boundary-bug analysis on a diff or branch. Auto-detects entry points, traces flows through changed code, finds every seam (cross-module calls, serialization, file I/O, shared state, schema versioning, network/IPC), and refuses to mark the work complete until each unverified boundary has a real round-trip test or an explicit written out-of-scope record persisted in an audit file. Use whenever the user says "boundary check", "seam check", "round-trip check", "flow boundaries", "user-flow check", "before merge", "is this safe to ship", "pre-merge gate", "boundary bugs", "verify the joins", or asks to validate cross-module joins, producer/consumer contracts, or end-to-end coverage of a change. Also use as a final gate from pr-review on any diff that touches more than one file, module, or process. Be pushy. Most surviving production bugs live at seams, not inside units — if the diff crosses any boundary, this skill almost certainly applies.
development
Run existing work through 5 specialist craftspeople who each produce an improved version, then peer-review and synthesize the best into a single improved artifact. Use when the user says "forge this", "improve this", "make this better", "level this up", "refine this", or asks for multi-angle improvement on code, copy, strategy, plans, designs, or any artifact where the current version works but could be significantly better. Do NOT use for decisions (use llm-council), simple edits, or creation from scratch.
development
Comprehensive test suite audit that combines ruthless analysis with a solution-focused roadmap. Reads test suites (unit, integration, e2e) and source code, produces a brutal audit report of test quality and gaps, and generates prioritized testing improvements.
development
Propose how to upgrade a resume by identifying missing projects, skills, evidence, and experience-building moves, but only after a `resume-critic` review has been completed for a specific job description or job title. Use when the user wants strategic suggestions beyond the current resume.