agentleFS
avibebuilderavibebuilder/claude-prime/.claude/skills/optimus-prime/SKILL.md

optimus-prime

Skill119 starsChanged 6 months agoMITView on GitHub

Generates a Claude Code configuration tailored to a specific project. Use whenever the user wants to prime a project, set up claude for a repo, bootstrap claude config, or re-prime/refresh an already-primed project (for example after `/prime-sync` pulled new starter content). Triggers on 'prime', 'prime this project', 'optimus-prime', 're-prime', 'refresh claude config', 'regenerate CLAUDE.md', 'set up claude for this repo'. Deeply analyzes the real codebase and builds project-specific skills, rules, and CLAUDE.md — not generic boilerplate. For ongoing config health checks and proposal review, use `self-evolve` instead.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 8 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Default to re-prime whenever meaningful existing config is found.

    line 19
  2. Distinctive:

    Follow full step-by-step in setup-project.md.

    line 33
  3. Distinctive:

    Every piece of knowledge must earn its place.

    line 37
  4. Distinctive:

    Repo evidence is the source of truth: Verify conventions from actual source files, configs, and docs.

    line 50
  5. Distinctive:

    Lean context, high signal: Follow the context engineering philosophy — load only what's needed, when it's needed.

    line 51
  6. Very specific:

    Rules are optional and path-scoped: Only create a rule when the rule test passes.

    line 52
  7. Distinctive:

    Reuse over duplication: If the project has strong docs, point to them.

    line 53
  8. Distinctive:

    Before making non-trivial changes, report findings in this shape:

    line 57
  9. Distinctive:

    After approval, implement and finish with a short summary of what changed and any follow-ups.

    line 72

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Ultrathink.

Mission

Generate a Claude configuration that fits the specific project. Every skill, rule, and CLAUDE.md entry must exist because this project needs it — not because a template included it. Analyze the real codebase deeply, understand its conventions and patterns, and build a config that makes Claude work well here.

Modes

  • Fresh prime — no claude config yet. Build from scratch.
  • Re-prime — claude config already exists. Preserve intentional project work, refresh what is stale or generic, fill gaps.

Default to re-prime whenever meaningful existing config is found. Never overwrite it wholesale without user approval.

Flow

  1. Analyze the repo deeply — stack, conventions, boundaries, docs, existing config. See analysis-checklist.md.
  2. Propose findings using the Output Format below. Do not touch meaningful files before approval.
  3. Create skills via /skill-creator — every skill must go through skill-creator to ensure quality and project fit. Starters in .claude/starter-skills/ are reference input to accelerate creation, not templates to copy.
  4. Generate or refine CLAUDE.md — lean, high-signal, always-on context only. Point to docs/READMEs for detail.
  5. Rules only if needed — apply the rule test. Zero rules is valid.
  6. Offer CLAUDE.local.md — personal preferences (role, sandbox URLs, preferred test data, workflow quirks), gitignored.
  7. Clean up — delete .claude/starter-skills/ after processing. Keep protected skills. Confirm all deletions.
  8. Verify — references resolve, stack claims match evidence, config is project-specific not generic.
  9. Offer skill optimization — after everything is set up, offer optimization paths sized to how many new skills were created. ≤3 skills: recommend full optimization for all. >3 skills: recommend full optimization for user-designated core skills only + description optimization for the rest (avoids long execution time and token burn). Always emphasize that full optimization takes meaningful time and tokens. Run sequentially — never in parallel; concurrent runs overload the user's machine.

Follow full step-by-step in setup-project.md.

Placement Decision Matrix

Every piece of knowledge must earn its place. Use this to decide where it belongs:

Detected needWhere it belongsTest
General framework/library knowledgeSkill (via /skill-creator)Would a reusable skill teach this across projects?
Project-specific constraint that still produces wrong code with the right skill loaded.claude/rules/<name>.md with paths:With the relevant skill activated, will code still be wrong without this?
Identity, commands, stack, key architecture, reference pointersCLAUDE.mdShould this be always-on for most tasks in this repo?
Detailed architecture or domain explanationsExisting docs/, READMEs — referenced from CLAUDE.mdValuable but too detailed for always-on?
Dense agent-oriented reference material - referenced from CLAUDE.md.claude/project/ (optional)Do agents need a tighter reference than the human docs provide?

Principles

  • Domain-scoped, project-fitted. Skills target a capability, not the project itself — with this project's patterns baked in. Generic templates and project-named catch-alls both degrade context quality.
  • Repo evidence is the source of truth. Verify conventions from actual source files, configs, and docs. Do not assert what you haven't confirmed.
  • Lean context, high signal. Follow the context engineering philosophy — load only what's needed, when it's needed. CLAUDE.md carries always-on essentials. Skills load on demand. Rules auto-attach by path.
  • Rules are optional and path-scoped. Only create a rule when the rule test passes. Do not modify _apply-all.md — it is a universal boilerplate rule from prime, not project config.
  • Reuse over duplication. If the project has strong docs, point to them. Do not re-author parallel agent docs unless existing material is too noisy or incomplete.

Output Format

Before making non-trivial changes, report findings in this shape:

Current State

What Claude config already exists; what the repo evidence says about stack, tooling, and conventions.

Proposed Changes

What to create, update, keep, or remove — including which skills to build and what each should cover.

Files to Touch

Concrete file list with short purpose notes.

Include these only when non-empty:

  • Decisions Needing User Input — overwrites, deletions, or major structural changes
  • Risks / Assumptions — inferred facts that still need verification

After approval, implement and finish with a short summary of what changed and any follow-ups.

References

ReferenceContent
analysis-checklist.mdWhat to inspect during repo + existing-config review
setup-project.mdFresh-prime / re-prime step-by-step workflow

Additional Context (Optional)

<user-context>$ARGUMENTS</user-context>

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.