Skip to content
Testing
Skill

/plan

Plan a phase or feature DIRECTLY into Beads — no PLAN.md, no markdown bridge. Runs GSD-style socratic discovery + decomposition + a goal-backward self-check, then emits a structured plan that `cmpl plan-apply` materializes as a Beads epic + worker-contract tasks + dependency

From plugin
completely
516 skills9 agents
Install
$ npx -y skills add 23ag1/completely --skill plan --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/plan

Context preview

The summary Claude sees to decide when to auto-load this skill.

Plan a phase or feature DIRECTLY into Beads — no PLAN.md, no markdown bridge. Runs GSD-style socratic discovery + decomposition + a goal-backward self-check, then emits a structured plan that `cmpl plan-apply` materializes as a Beads epic + worker-contract tasks + dependency

SKILL.md

plan.SKILL.md
name: completely:plan
description: Plan a phase or feature DIRECTLY into Beads — no PLAN.md, no markdown bridge. Runs GSD-style socratic discovery + decomposition + a goal-backward self-check, then emits a structured plan that `cmpl plan-apply` materializes as a Beads epic + worker-contract tasks + dependency waves + human checkpoints. One source of truth from the first planning act. Use to turn an idea or phase into ready Beads work.
version: 0.4.0
user-invocable: true
argument-hint: "[phase or feature]"

Plan **into Beads directly**. Do NOT write a `PLAN.md` / `STATE.md` — the plan lives in Beads.

Method (GSD methodology, Beads-first)

1. **Discover (socratic).** Ask clarifying questions ONE at a time until the goal is unambiguous; offer 2–3 approaches with trade-offs; confirm scope. If a spec is already frozen, read it instead. 2. **Decompose** into atomic tasks (2–5 min each). Each task MUST carry the worker-contract: `title` (imperative), `acceptance` (observable/testable), `design` (approach/why), `write_zone` (exact files/globs), `verify` (the proof command), `deps` (other task keys). ~2–3 tasks per unit; split if a unit exceeds ~5 tasks or ~10 files (context budget). **Also fill the context fields — the engine consumes them directly:**

  • `read_context` — the exact files/interfaces the worker must READ first but won't edit

(contracts, schemas, neighbouring patterns). Injected verbatim into the worker's dispatch header, so a fresh context reads exactly this instead of guessing or scanning the tree.

  • `must_haves` — the goal-backward checklist `{truths, artifacts, key_links}` the default-FAIL

evaluator grades against at VERIFY. No must_haves ⇒ the evaluator has only the acceptance line.

  • `requirements` — requirement IDs this task covers (traceability), when the phase has them.

3. **Checkpoints.** Where a human must verify/decide, add `{key,title,after,how}`; make downstream tasks depend on `cp:<key>` so the wave blocks until the human closes it. 4. **Self-check (goal-backward, GSD plan-checker's 7 dimensions).** Before applying, verify: requirement coverage · task completeness (acceptance+verify present) · dependencies acyclic · artifacts wired together (not isolated) · scope within context budget · acceptance is user-observable · honors the user's frozen decisions. Fix gaps — don't apply a plan that won't reach the goal. 5. **Apply.** Emit the plan as JSON and pipe it to the materializer: `cat <<'PLAN' | cmpl plan-apply` … `PLAN`. It creates the epic + tasks + deps + checkpoints + a `bd swarm` (native waves). Then show `bd ready` and `bd swarm status`.

Plan JSON schema

`{ "epic": "...", "tasks": [ {"key","title","acceptance","design","write_zone":[...],"verify","deps":[...],"labels":[...],"read_context":[...],"must_haves":{"truths":[...],"artifacts":[...],"key_links":[...]},"requirements":[...]} ], "checkpoints": [ {"key","title","after","how"} ] }`

Rules

  • Beads is the only artifact. Never persist a markdown plan.
  • Every task lint-clean (acceptance + design + write_zone) — `plan-apply` + `cmpl lint` enforce it.
  • One question at a time; respect a frozen spec; don't over-interrogate.
Read more
Ships withcompletely

A quality-first harness for autonomous AI coding agents. It turns "the agent said it's done" into *"here's the proof — graded by an independent, default-FAIL checker."* Done is earned, not asserted.

Get the whole plugin

Other skills on completely.

auto
Skill

auto

Run completely AUTONOMOUSLY — loop the FULL task engine over the Beads queue, a fresh `claude -p` per task, until the queue is empty. Same complete recipe as…

check
Skill

check

Run all configured quality checks (lint, types, tests) in one pass with terse output — reports "clean" when green, and only the failing check's output when…

control
Skill

control

Run completely UNDER CONTROL — execute the SINGLE next Beads task through the FULL task engine (understand → map → plan-check → parallel subagents → TDD →…

init
Skill

init

Scaffold the completely thin layer into the current repository — Definition of Done, the harness rules snippet in CLAUDE.md, and an optional project-specific…

run
Skill

run

Drive the Beads queue (`bd ready`) with one engine in two autonomy modes — supervised (GSD wave subagents, human gates at phase boundaries) or unattended…