opentabs / bump-version
CLAUDE.md926 starsChanged 8 months agoMITView on GitHub
AI clients load SKILL.md once at session start and cache its frontmatter + content in memory. To ensure the AI always gets the latest skill instructions (even if the skill was updated after the session started), SKILL.md is a lightweight stub that tells the AI to read __SKILL__.md at execution time. When you need to update the bump-version skill content: Update __SKILL__.md when: Some AI clients (e.g., Claude Code with the skill tool) load skill files into memory once at…
In this repository17 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .agents/skills/build-plugin/CLAUDE.mdCLAUDE.md
- .agents/skills/bump-version/CLAUDE.mdCLAUDE.md
- .agents/skills/ralph/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- .claude/skills/build-plugin/CLAUDE.mdCLAUDE.md
- .claude/skills/ralph/CLAUDE.mdCLAUDE.md
- docs/CLAUDE.mdCLAUDE.md
- e2e/CLAUDE.mdCLAUDE.md
- plugins/CLAUDE.mdCLAUDE.md
- .ralph/CLAUDE.mdCLAUDE.md
- build-pluginSkill
- bump-versionSkill
- ralphSkill
- build-pluginSkill
- bump-versionSkill
- ralphSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 21
When you need to update the bump-version skill content:
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Bump Version Skill — Maintenance Guide
File Structure
.claude/skills/bump-version/
├── SKILL.md # Stub with frontmatter (loaded by AI clients at session start)
├── __SKILL__.md # Actual skill content (read on-demand for latest version)
└── CLAUDE.md # This file — maintenance instructionsHow It Works
AI clients load SKILL.md once at session start and cache its frontmatter + content in memory. To ensure the AI always gets the latest skill instructions (even if the skill was updated after the session started), SKILL.md is a lightweight stub that tells the AI to read __SKILL__.md at execution time.
SKILL.md: Contains only the YAML frontmatter (name, description, triggers) and an instruction to read__SKILL__.md. Do NOT put actual skill logic here.__SKILL__.md: Contains the full skill instructions — the version bump workflow, file inventory, and verification steps. This is the file to edit when updating the skill.
Updating the Skill
When you need to update the bump-version skill content:
- Edit
__SKILL__.md— this is the single source of truth for skill behavior - Do NOT edit
SKILL.mdunless you need to change the skill's name, description, or trigger keywords in the frontmatter - Changes to
__SKILL__.mdtake effect immediately for any AI session that invokes the skill, since the AI reads it fresh each time
When to Update This Skill
Update __SKILL__.md when:
- A new platform package is added to
platform/ - A new plugin is added to
plugins/ - The dependency graph between packages changes
- The version referencing pattern changes (e.g., if plugins switch from
^x.y.ztoworkspace:protocol) - New files with hardcoded versions are discovered
Why This Pattern Exists
Some AI clients (e.g., Claude Code with the skill tool) load skill files into memory once at the start of a session. If the skill content is updated mid-session, the AI continues using the stale cached version. By keeping SKILL.md as a stub that redirects to __SKILL__.md, the actual content is read on-demand via the Read tool, guaranteeing the latest version is always used.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

