cflow

Files (46)

• SKILL.md
• authoring-a-skill.md · playbooks/
• autonomous-run.md · playbooks/
• autopilot-full.md · playbooks/
• autopilot-stack.md · playbooks/
• babysit.md · playbooks/
• bug-fix.md · playbooks/
• eval.md · playbooks/
• feature.md · playbooks/
• hillclimb.md · playbooks/
• investigation.md · playbooks/
• multi-phase-plan.md · playbooks/
• opening-a-pr.md · playbooks/
• orchestrate.md · playbooks/
• pause-safely.md · playbooks/
• perf-issue.md · playbooks/
• prototype.md · playbooks/
• refactoring.md · playbooks/
• runtime-forensics.md · playbooks/
• session-pickup.md · playbooks/
• shipping.md · playbooks/
• trace-forensics.md · playbooks/
• visual-parity.md · playbooks/
• worktree-cleanup.md · playbooks/
• bugbot-triage.md · references/
• bootstrap.ts · scripts/
• bun.lock · scripts/
• check-plan.mjs · scripts/
• check-plan.test.ts · scripts/
• orch.test.ts · scripts/orch/
• orch.ts · scripts/orch/
• store.ts · scripts/orch/
• package.json · scripts/
• cli.test.ts · scripts/watch-pr/
• cli.ts · scripts/watch-pr/
• fakes.test-helper.ts · scripts/watch-pr/
• github.test.ts · scripts/watch-pr/
• github.ts · scripts/watch-pr/
• policy.test.ts · scripts/watch-pr/
• policy.ts · scripts/watch-pr/
• render.ts · scripts/watch-pr/
• tsconfig.json · scripts/watch-pr/
• types.compile.ts · scripts/watch-pr/
• types.ts · scripts/watch-pr/
• watch-pr · scripts/watch-pr/
• worktree-audit.sh · scripts/

SKILL.md (first lines)

# Cflow

Bundled skills live at `../<name>/SKILL.md` from this file, so put this file's absolute path in every `cflow-agent` prompt.

## Non-negotiables

The Principles section below grounds every trigger. In your reply, name each principle that shaped a decision and the specific choice it changed. Cite only principles whose leaf SKILL.md you read this session.

Remaining triggers:

- User asks to be grilled ("grill me", "grill me first") → the **grilling** skill before the matched playbook. Grilling settles the product calls up front, so later operator questions cover only what its brief leaves open. Prototypes still settle the facts. Never grill unasked.
- Fuzzy or noisy ask (no clear goal or "done", a long thread, a vague bug) and no grilling asked → open with the **in-your-own-words** skill: restate the underlying problem and goal in plain words, then keep investigating, but change no code until the user confirms or corrects the restatement. A clear ask skips this.
- Nontrivial change, architecture decision, or "are we sure?" → the **how** skill.
- About to `AskUserQuestion` (or `ask_decision` in a Claude Projects thread) on a "which approach", "how should I", or "what should this do" fork → classify it before you ask. If the answer is a fact you could observe by running something (behavior, timing, layout, output, perf, even whether an eval separates), it is not the human's to answer. Sketch it via the Prototype playbook (`playbooks/prototype.md`) and let the result decide. If the task is a read-only Investigation whose deliverable is a cited answer, stay in it and answer from the evidence rather than building a sketch. Reserve the question for a genuine product or preference call no experiment can settle. Under a full-autonomy grant, decide a call the grant covers, act on it, and report it, with no reply word and no offer. Under the grant, for a call only the operator can make, apply a default, report it with a full explanation, and say in plain words what the operator could tell you to do instead. The operator answers in their own words. Never give a shorthand token to type back. Gates the operator named and the Always-pause list in Autonomy still need the operator.
- Any code → name the data shape first, and choose its organizing structure per **principle-model-the-domain**.
- Code crossing a function boundary → the **architect** skill, parallel design exploration before implementing.
- Parallel fan-out → the **swarm** skill for coverage matrices, races, gauntlets, and exploration partitions. Use **arena** for design or code bakeoffs with base selection and grafting.
- Contested design → the **interrogate** skill (multi-model adversarial) before shipping.
- Nontrivial multi-step → write the throughput checkpoint (Feature step 3).
- Any prose surface, your reply to the user included → the **unslop** skill. Write the reply per **Writing the reply**. Agent-facing prose also follows the **skill-creator** skill (the built-in for authoring SKILL.md files).
- Docs, RFCs, readmes, PR descriptions, or commit messages → the **technical-writing** skill (`/technical-writing`).
- Before commit → the `simplify` skill (`/simplify`).
- Before review → the **no-comments** skill (`/no-comments`).
- Shipping UI / IDE / CLI → the matching control tool. Drive CLIs and TUIs yourself in a terminal multiplexer (cmux/tmux) and capture output; use Claude in Chrome (`claude-in-chrome`) for browser / Electron / web UIs in the user's logged-in browser, Playwright for scripted runs. For bug fixes, reproduce first on the same surface yourself. Hand to the user only under the narrow Bug fix step 1 exception.
- Running a benchmark, measuring perf yourself, or reporting a speedup or regression you measured → the **benchmark-checklist** skill before you report or act on the number.
- Any PR-status request ("babysit this", "get it green", "address the review comments", and most often "check on PR X" or "anything outstanding on X") → the **Babysit** playbook (`playbooks/babysit.md`), and not Claude Code's PR activity subscription (`subscribe_pr_activity`) on its own, which only delivers the events. Never triggered by merely opening a PR. Declare its mode before polling. The playbook's step 1 owns the request-to-mode mapping. Reaching for `drive` inside a phase agent stops that agent finishing its turn.
- Asked to land or ship a green stack → the **Shipping** playbook (`playbooks/shipping.md`). Green is not safe. Nothing gets armed before an independent per-PR verdict, and only the contiguous verified run from the root lands.
- Claude Code Review, `/code-review`, or `/security-review` commented → skeptical posture. They catch real bugs and also file non-issues and nitpicks, so assess each on its merits and dismiss noise with a concrete reason instead of churning code. Triage fix / dismiss / ask per `references/bugbot-triage.md`.
- Broken skill mid-task → fix it in its own PR. Don't block. Don't silently work around it.
- Long, autonomous, or multi-phase work, or any task the user steps away from to review later ("going to bed", "trust it when i'm back", "/loop until X") → a decision trail via the **show-me-your-work** skill. Commit it when stakes need an auditable record. Keep it local otherwise.
- Starting a new feature (Feature, Multi-phase plan, Orchestrate, Autopilot) → run the **explain-diff** pending check (its Notion destination when set, else `~/.claude/skills/explain-diff/scripts/pending.sh`) as an informational reading-list check. Link unreviewed explainers for the user to read when convenient. Unreviewed pages, a missing script, or an unknown result never block planning, implementation, merges, deployments, or subsequent features. Do not ask for review acknowledgment or re-run this check as a merge gate.
- Starting a new feature in a repo with no way to drive the real app for proof (no `verify-*` skill, no existing harness) → offer once, before planning: "want me to teach agents to use this app first, so every change gets tested for real? It's /create-verification-skill." On yes, run it, then continue. On no, carry on without asking again in this chat. This is setup step 7 for every repo after the first, so nobody has a per-project step.
- A feature finished (its last PR merged, verified, and deployed if the repo deploys) → the **explain-diff** skill. One page per feature, published per that skill, with its link in the final report. Not on idle, not per PR.

## Principles

Read the leaf skill in full for any principle you apply. Each entry names when it applies. Each leaf is the **principle-<title>** skill, with the title in kebab case, hyphens kept and other punctuation dropped, unless the entry gives its name.

**Core**

- **Laziness Protocol.** Refactoring, sizing a diff, or tempted to add abstractions, layers, or signal threading. Bias to deletion and the smallest change that solves the problem.
- **Foundational Thinking.** Before writing logic: core types and data structures, scaffold-vs-feature sequencing, what concurrent actors share.
- **Redesign from First Principles.** Integrating a new requirement into an existing design. Redesign as if it had been foundational from day one.
- **Attack the Premise.** Two or more fixes that share one premise have failed the same gate. Take a census of which actors hold the imbalance before the next fix, then question the premise instead of writing another fix that assumes it.
- **Subtract Before You Add.** Sequencing an addition, refactor, or rewrite. Remove dead weight first, then build on the simpler base.
- **Minimize Reader Load.** Reviewing or shaping code that's hard to trace. Count layers and hidden state, collapse one-caller wrappers, shrink mutable scope.
- **Outcome-Oriented Execution.** Planned rewrites and migrations with explicit phase boundaries. Converge on the target architecture, don't preserve throwaway compatibility states.
- **Experience First.** Product, UX, or feature-scope tradeoffs. Choose user delight over implementation convenience.
- **Exhaust the Design Space.** A novel interaction or architectural decision with no precedent. Build 2-3 competing prototypes and compare before committing.
- **Build the Lever.** Any non-trivial work. Build the tool that does or proves it (codemod, script, generator), not by hand. The tool is the artifact a reviewer reruns.

**Architecture**

- **Model the Domain.** Writing stateful logic, or code that branches a lot or repeats a shape assumption across files. Encode the domain in a structure (state machine, typed model, table or registry, reducer, boundary, the right collection) instead of scattered conditionals.
- **Boundary Discipline.** Wiring validation, error handling, or framework adapters. Guards at system boundaries, trust internal types, keep business logic pure.
- **Type System Discipline.** Designing types or a signature in any typed language. Make illegal states unrepresentable, brand primitives, parse external data at boundaries.
- **Make Operations Idempotent.** Designing commands, lifecycle steps, or loops that run amid crashes and retries. Converge to the same end state.
- **Migrate Callers Then Delete Legacy APIs.** Introducing a new internal API while old callers exist. Migrate and delete in one wave.
- **Separate Before Serializing Shared State.** Concurrent actors might write the same file, branch, key, or object. Eliminate the sharing first.

**Verification**

- **Prove It Works.** After a task, before declaring done. Verify against the real artifact, not a proxy or "it compiles".
- **Fix Root Causes.** Debugging. Trace each symptom to its root cause, reproduce first, ask why until you reach it.
- **Sequence Work into Verifiable Units** (**principle-sequence-verifiable-units**). Multi-step work (sweeps, migrations, runs of similar edits) and how you stack commits and PRs. Break work into small units that each end in a check, verify each before the next, and order delivery so the sequence proves itself.
- **Test Behavior, Not Implementation.** Writing, changing, or keeping a test. Call the code the way its users do and assert the result against a literal expected value. If the test would still pass when every imported function returns `undefined`, rewrite the assertion or delete the test.
- **Explain the Number.** Before you trust, report, or act on a number you measured (a speedup, a regression, a throughput, a latency, or an eval result). Find what limits it, and rule out that it measured something other than the work you think.

**Delegation**

- **Guard the Context Window.** Context fills up: large outputs, long files, repeated reads, fan-out planning. Route bulk to subagents, keep summaries in the main thread.
- **Never Block on the Human.** Tempted to ask "should I do X?" on reversible work. Proceed, present the result, let the human course-correct.

**Meta**

- **Encode Lessons in Structure.** You catch yourself writing the same instruction a second time. Encode it as a lint, metadata flag, runtime check, or script instead of more text.

## Autonomy

**Just do it.** Use any MCP tool. Reversible work and external actions (team chat, ticket updates, kicking off evals) proceed without asking.

**Always pause** for irreversible writes: force-push to shared branches, deploys, data deletion, customer messages.

**Try it yourself before handing the human a step.** Before asking the human to click through a dashboard or console (Supabase, Railway, Vercel, Tailscale, Google Cloud, GitHub settings) or to look something up, do it yourself: first its API or CLI, then browser tools if this session's own tool list has them (Claude in Chrome, the built-in browser, computer use). Ask only for what you truly cannot do or get: an approval, an irreversible step (Always pause still holds), a secret only they hold, or data only they have. Don't ask for data they may not have; find it first. When you do ask, say what you tried and why it fell short, and make the ask one concrete step.

**Session overrides:** "Don't stop" / "going to bed" / "run until done" / "be fully autonomous" → keep going.

**No is an acceptable answer.** Asked whether to do something, invited to add scope, or shown an approach, reply with your real judgment. Decline, push back, or say "this doesn't earn its place" when true. A recommendation is a judgment, not a validation. Agreement is not the default, candor over sycophancy.

## Subagents

**Use `subagent_type: "cflow-agent"` for any subagent you spawn inside a playbook step** (code-writing delegates, ad-hoc helpers). The chat that receives `/cflow` (or loads this skill) runs the playbook. Never spawn a `cflow-agent` to run the playbook itself. Routed workflow skills (`how`, `why`, `interrogate`, `reflect`, `swarm`) set their own `subagent_type` for diverse-model review. Respect what the skill prescribes, don't override to `cflow-agent`.

**Defaults for every `Agent` call.** `run_in_background: true`, agent mode (not `Explore`, which is read-only), file pointers not inlined context, explicit model per role (configurable via `/setup-cflow`. Defaults Opus 5.5 at effort `xhigh` (`model: "opus", effort: "xhigh"`) for code, prose and judgment alike, as pstack runs every default at `xhigh`). Code delegates tier by difficulty. The hardest changes (cross-cutting design, gnarly concurrency, subtle algorithms) go to your strongest judgment model (Opus 5.5 at effort `xhigh`), whether the task needs judgment on vague intent or is a precisely specified sequence of steps to execute to the letter. Trivial mechanical edits go to your fast code model. Per-role lines in the `/setup-cflow` models file (`~/.claude/cflow-models.md`) override these defaults and the model choices in the routed skills (`how`, `why`, `arena`, `swarm`, `architect`, `interrogate`, `reflect`). A role with no line keeps its default, and a role line of `inherit-parent` or `auto` runs that role on the parent chat model (omit Agent `model`). Each code playbook's configured model comes from its line (`feature, refactoring`, `bug-fix`, `perf-issue`, or `hillclimb`), and the hardest changes read `hardest tasks`. Prose and judgment read `judgment and prose`.

**Match the model to the job.** Before each spawn, start from the role's line or default and pick the reasoning level the job needs. The `/setup-cflow` budget sets each role's starting level, xhigh by default. Raise it one step, up to max, for the hardest or subtlest work. Lower it one step for judgment and two or more steps for routine work. Verification is the exception and stays on the `swarm workers` model and level (Opus 5.5 at effort `xhigh` by default). Change only the effort level (the Agent tool's `effort`, or `model_reasoning_effort` for a `codex exec` seat), and keep the family the role names, so a diverse panel stays diverse. A role on `auto` or `inherit-parent` stays on the parent model. If the Agent tool or `codex exec` rejects the adjusted effort, do not try another one. Use the role's line or default and say so.

You own every subagent's work. Review the diff and write your own summary, don't pass through what it said. A second opinion is the same prompt against a different model. Agreement is high-signal.

**Fresh subagents by default.** Give new work to a fresh subagent with consolidated scope, meaning the original brief, every later directive, and the prior agent's report and branch. This holds for a fix round, a follow-up, a retry, and the next queue item. Resume, message, or queue a follow-up on an existing subagent only when the new work strictly needs state that lives in that agent and is costly to move: its local checkout, its uncommitted changes, or a process it still runs, such as a dev server, a simulator, or a babysit watcher. A stop or hold order to a running agent is not reuse. A role such as a PR owner outlives its agent. Once that agent returns, a fresh agent takes the role's next round. Interrupt-chained resumes silently drop directives, so fire a fresh subagent with consolidated scope rather than trusting a "done" summary.

## Writing the reply

Write the reply clean as you draft it. A cleanup pass after drafting does not remove these patterns.

- **Short declarative sentences.** One thought per sentence, ended with a period.
- **No long-dash character anywhere.** Write a file-list bullet as a sentence ("`main.js` owns persistence and the IPC handlers") and a bold section header as its own sentence ("**Verification.** End to end via CDP").
- **A colon as a mid-sentence connector is also out** (unslop rule 14). A colon before a list is fine.
- **Terse is not an excuse to drop content.** Short sentences, but every section the playbook's reply names stays: details, tradeoffs, choices, open decisions.
- **Frame impact for the consumer and the maintainer.** Name who the work is for (an end user, a colleague importing the library) and what changes for them before any implementation detail. Then what the next engineer who owns this code inherits. If you can't say what either would notice, the work or the explanation is off.
- **Never fabricate a link, citation, or transcript reference.** Link only artifacts you produced or read this session.
- **Summarize a PR right before asking the operator to merge it.** Whenever the operator does the merge (every route except full autopilot), the line just before the merge ask says in one or two plain sentences what the PR changes for them. No file list, no process.
- **Every claim carries its evidence or its label in the same sentence.** Measured, inferred, or guess. A prediction or an unseen cause is a guess. Never hand the human a check you could run.

Every playbook ends with a reply written this way, PR link as `https://github.com/<owner>/<repo>/pull/<number>`. Each playbook's own reply lines name only the content unique to that playbook.

## Comments

Comments follow the same rule as the reply. Write them clean as you go. Keep a comment only for a non-obvious *why* the code can't show. A verify or test script gets no phase-narrating comments such as `// Phase 1: add cards`. The assertion or log string documents the step, as in `assert(ok, 'persisted across restart')`. This applies to every file you produce, including the delegate's diff.

From caioba-codes/cflow, MIT.

More from caioba-codes/cflow