agentleFS
Sign inStart free
ruverdruverd/bulma/skills/bulma-feature-delivery/SKILL.md

bulma-feature-delivery

Skill1 starsChanged 50 days agoMITView on GitHub

Use when delivering a ticket, bug, or chore end-to-end (implementation through draft PR with CI green) via /bulma-feature-delivery or /bulma-fd, or when resuming a run whose HANDOFF.md exists. Spine: grill -> spec -> tickets -> implement(tdd) -> review.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    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
  • could be automatic: a setting could enforce it
  1. Very specific:

    Never merge. Draft PR only from shipper.

    line 61
  2. Distinctive:

    Before final evidence or a read-only report, check the selected playbook's proof.

    line 90
  3. Distinctive:

    Keep its identity and one checklist in STATE.

    line 91

≈ “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

Bulma feature delivery

Orchestrator. Chat: bulma-memory. Kevin. No product code on the main thread.

Load now:

  • GRAPH.md
  • STATE.schema.md
  • ROUTING.md
  • PLAYBOOKS.md
  • PRODUCT.md
  • VOICE.md
  • PSTACK.md
  • DECISION_POLICY.md
  • bulma-bus DISK.md
  • bulma-memory

Spine (bundled primitives):

bounded discovery → contract → implement(/tdd) → code-review

Bounded discovery: GRILL.md. Interview skills are optional for user-requested design exploration. Ask only under DECISION_POLICY.md.

Quick start

/bulma-fd ABC-123
/bulma-fd ABC-123: extra note
/bulma-fd login button does nothing
/bulma-fd resume
/bulma-fd … --no-pr

Checkout is feature/<id-lowercase>. Worktree and branch rules: JOBS.md §Worktree.

Orchestrator loop

Resume: read STATE + HANDOFF, RECONCILE (replay HANDOFF invariants; git/gh vs STATE). Continue at the current node. Skip a settled node only when its invariants match. If waiting_user, the user message is the ASK answer. Do not re-init. Do not re-grill settled decisions unless an AC line contradicts them. Missing risk on a v4 STATE → normal.

Fresh:

  1. Init STATE under $BULMA_ROOT/.bulma-feature-delivery/. Load bulma-memory.
  2. Recognize explicit read-only or planning-only intent before mcp_context. It disables checkout creation. Then mcp_context and triage select the task contract under PLAYBOOKS.md. Critical MCP down → STOP.
  3. Walk GRAPH. Grill / spec / tickets on the main thread. plan_critic on the main thread unless risk=elevated (then one read-only spawn). Spawn one node subagent at a time for implement / review / diagnose / tester / evidence / quality / shipper.
  4. Never merge. Draft PR only from shipper.
  5. Near context limit → handoff.

Paths

path / scopeWhenHow
full_featurenew behaviorgrill → spec → tickets → plan_critic (skip if risk=low) → TDD implement
debug_fixbugdiagnose → one TDD ticket
light_changechoreone ticket → one coder
scope: fullstackFE and BE, sibling resolvedsame branch, git worktrees (FULLSTACK.md; Orca optional)

Ship: implement → blast when required → quality → review → tester → final evidence → commit → push → draft PR (reviewers and assignee: PRODUCT.md) → CI 100% green (only then delivered).

Intelligence

  1. Route with evidence. Not full_feature on every bug.
  2. DECIDE micro-choices. ASK only last-resort policy.
  3. Bug: root cause before the fix. If it is a feature, re-route to grill.
  4. Implementation only in a subagent. Main does not edit product code.
  5. TDD on every behavior change. Quality before independent review; deep thermo on elevated or structural changes.
  6. UI: DS of the repo; Figma if present; else copy recent same-type screens.
  7. Discover the repo first (PRODUCT.md). Critical tracker URL offline → error and stop. Local goal does not need a tracker.
  8. Blocker: advance what you can → Draft a tracker issue + contract comment → waiting_blocker.
  9. Tokens: TOKEN_ECONOMY.md.
  10. Limit near: HANDOFF.md.

Before final evidence or a read-only report, check the selected playbook's proof. Keep its identity and one checklist in STATE. Resume reconciles existing rows; it never fabricates a baseline or restarts completed work.

Files

Load on demand. This list is the index.

bulma-feature-delivery/
├── SKILL.md GRAPH.md ROUTING.md PSTACK.md GRILL.md VOICE.md
├── DECISION_POLICY.md TDD.md IMPLEMENTATION.md
├── MCP_CONTEXT.md LINEAR.md BLOCKERS.md FULLSTACK.md PRODUCT.md
├── UI_DESIGN_SYSTEM.md CI_DELIVERY.md HANDOFF.md TOKEN_ECONOMY.md
├── STATE.schema.md
├── nodes/   templates/   adapters/

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.

Check your setup, free