agentleFS
lglucaslglucas/ai-dev-operating-system/.claude/skills/registry-pick/SKILL.md

registry-pick

Skill12 starsChanged 47 days agoMITView on GitHub

Recommend external repo packs from docs/registry/ that fit a project's stack, domain, and constraints. Runs twice in the WIZARD — a design-scoped pass at stage 3.1 (before the Prototype Lab) and a full stack pass at stage 4.3 (after the Technical Plan). Also use when the user asks "que repos eu deveria usar?", or whenever a new need (security review, design system, deploy guidance) appears mid-build.

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. About 0 are common, 7 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read docs/registry/INDEX.md and docs/registry/packs/*.md, match the project's signals against pack tags and fit-signals, and produce a …

    line 8
  2. Distinctive:

    When the user asks "que repos eu uso aqui?", "tem algum pack pra X?", "que skill externa ajuda nisso?".

    line 29
  3. Distinctive:

    When a new domain need surfaces mid-build (compliance, design system, malware analysis, web3 integration).

    line 30
  4. Distinctive:

    After adding a new pack to the registry, to retest the recommendation for the current project.

    line 31
  5. Distinctive:

    Before recommending, gather these from the project:

    line 37
  6. Distinctive:

    If signals are missing, ask 2–3 targeted questions BEFORE recommending.

    line 48
  7. Very specific:

    Never invent packs: Only recommend slugs that exist in docs/registry/INDEX.md.

    line 107
  8. Very specific:

    Cite the one-pager: Every recommendation links back to docs/registry/packs/<slug>.md.

    line 109
  9. Very specific:

    Flag missing license info: If a pack has License: check upstream, surface that as an open question.

    line 111
  10. Distinctive:

    Re-run after changes: If the Technical Plan or Product Brief changes, re-invoke the skill.

    line 112
  11. Distinctive:

    When the user says "encontrei um repo novo, adiciona aí":

    line 119
  12. Very specific:

    Create docs/registry/packs/<slug>.md from the template in docs/registry/README.md.

    line 124
  13. Very specific:

    Add a row to docs/registry/INDEX.md.

    line 125
  14. Very specific:

    Add the slug to relevant docs/registry/tags/<tag>.md files.

    line 126
  15. Very specific:

    If the pack also influenced this OS itself, cross-link in UPSTREAM-SOURCES.md.

    line 127

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Registry Pick

Read docs/registry/INDEX.md and docs/registry/packs/*.md, match the project's signals against pack tags and fit-signals, and produce a short, opinionated recommendation.

This skill does NOT install anything. It shortlists what is worth installing for THIS project.


When to invoke

The wizard invokes this skill twice, in two differently-scoped passes:

PassWIZARD stageScopeOutput
Design3.1 — before the Prototype LabOnly design-oriented packs: UI systems, design tokens, component libraries, diagramming, screenshot/asset toolingdocs/technical/registry-pick-design.md
Stack4.3 — after the Technical PlanThe whole catalogdocs/technical/registry-pick.md

The split is deliberate: design packs are worthless once the prototype is already built, and stack packs are guesswork before the Technical Plan exists.

Signals available differ between passes. In the design pass there is no Product Brief, no Technical Plan, and no prototype-lab/ yet — read BP v0.0.2 and knowledge-base/ instead, and do not ask for artifacts that Phase 3 has not produced.

Also invoke:

  • When the user asks "que repos eu uso aqui?", "tem algum pack pra X?", "que skill externa ajuda nisso?".
  • When a new domain need surfaces mid-build (compliance, design system, malware analysis, web3 integration).
  • After adding a new pack to the registry, to retest the recommendation for the current project.

Inputs the skill needs

Before recommending, gather these from the project:

SignalSourceExample
Stackpackage.json, Anchor.toml, requirements.txt, Gemfile, Cargo.toml@solana/web3.js → web3/Solana
Domaindocs/business/BUSINESS-PLAN.md, docs/product/PRODUCT-BRIEF.md (stack pass only)"PHI" → healthcare, "pagamentos" → fintech
Compliancedocs/technical/TECHNICAL-PLAN.md security/privacy section (stack pass only); BP v0.0.2 risk section in the design passLGPD, GDPR, HIPAA, PCI
UI surfaceBP v0.0.2 personas and positioning in the design pass; Product Brief + prototype-lab/ in the stack passmobile, dashboard, B2C web
Public/privaterepo metadata, deploy intentpublic GitHub → enables gitleaks
Team familiarityuser statement"primeira vez com Claude Code" → start with #foundations only

If signals are missing, ask 2–3 targeted questions BEFORE recommending. Never invent stack details.


Procedure

  1. Read docs/registry/INDEX.md to enumerate available packs.
  2. Score each pack against the project signals (stack match, domain match, compliance match, team-readiness fit).
  3. Filter out packs whose When NOT to install matches the project shape.
  4. Resolve conflicts using the Conflicts and overlaps section in each one-pager (pick one of overlapping packs based on context).
  5. Order by priority: must-install → recommended → optional → skip-for-now.
  6. Output a short recommendation table + rationale + install commands copied from each pack one-pager.
  7. Defer install execution to the user (this skill never runs git clone / npm install itself).

Output format

## Registry recommendation for <project name>

Project signals detected:
- Stack: <e.g. Next.js + Supabase>
- Domain: <e.g. SaaS B2B compliance>
- Compliance: <e.g. LGPD>
- UI: <e.g. dashboard + landing>
- Public repo: <yes/no>

### Must install

| Pack | Why for this project | How |
|------|----------------------|-----|
| `<slug>` | <one sentence> | `<install command>` |

### Recommended

| Pack | Why | How |
|------|-----|-----|
| ... | ... | ... |

### Optional

| Pack | Why | How |
|------|-----|-----|
| ... | ... | ... |

### Skipped (and why)

- `<slug>` — <reason from `When NOT to install`>

### Open questions before installing

- ...

Rules

  • Never invent packs. Only recommend slugs that exist in docs/registry/INDEX.md.
  • Never install. Show the user the install command from the pack one-pager and stop.
  • Cite the one-pager. Every recommendation links back to docs/registry/packs/<slug>.md.
  • Never recommend archived packs.
  • Flag missing license info. If a pack has License: check upstream, surface that as an open question.
  • Re-run after changes. If the Technical Plan or Product Brief changes, re-invoke the skill.
  • Respect the pass scope. In the design pass (stage 3.1), recommend only design-oriented packs and do not ask for the Product Brief, the Technical Plan, or the prototype — none of them exist yet.

Adding new packs

When the user says "encontrei um repo novo, adiciona aí":

  1. Open the upstream URL.
  2. Confirm license, status, and active maintenance.
  3. Pick a slug (lowercase, hyphenated).
  4. Create docs/registry/packs/<slug>.md from the template in docs/registry/README.md.
  5. Add a row to docs/registry/INDEX.md.
  6. Add the slug to relevant docs/registry/tags/<tag>.md files.
  7. If the pack also influenced this OS itself, cross-link in UPSTREAM-SOURCES.md.
  8. Re-run this skill against the current project to see if the new pack changes the recommendation.

Cross-references

  • Registry root: docs/registry/README.md
  • Master table: docs/registry/INDEX.md
  • Tags: docs/registry/tags/
  • Origin/audit map: UPSTREAM-SOURCES.md

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.