arbeitsplan

Plugin arbeitsplan from marketplace werkstoff in Anselmoo/werkstoff. Commands read from the repo's manifest — 11 days ago.

Skills (10)

• arbeitsplan-backend

  Decides whether a compiled arbeitsplan run executes in-session, as a headless matrix of fresh claude -p processes, or inside the Workflow tool, and emits the checked backend object workflow.json requires. Use when the user asks whether to run something as a workflow or a matrix, whether a job should

• arbeitsplan-compile

  Turns a stated problem into an executable, budgeted workflow spec plus the takt beats that enforce its order. Use when the user describes a change and wants it run as an agentic workflow, asks to plan or set up a swarm, says the advice they have is too vague to act on, or asks what shape a piece of

• arbeitsplan-design

  Designs a workflow node by node before anything runs — for each step, what it does, where it runs (primary checkout, agent worktree, scratch directory), when (dependencies, human gates between runs) and how (kind, model, write scope, output schema, and for script nodes the exact command and its tool

• arbeitsplan-matrix

  Compiles and runs a headless sweep — cases x models x plugin-states x repeats, one fresh `claude -p` process per cell — after proving the environment can authenticate. Use when the user wants genuinely independent candidates, a real per-candidate --model, a plugin-present versus plugin-absent compar

• arbeitsplan-patterns

  Says which agentic pattern fits a problem and with what numbers — fan-out width, model tier, stop rule, cost — grounded in a frozen catalog of measured configurations. Use when the user asks what shape work should be run in, how many agents to use, whether to run something in parallel, or whether a

• arbeitsplan-preflight

  Read-only inventory of what a swarm run can and cannot do in this repository before any tokens are spent. Use before arbeitsplan-compile or arbeitsplan-run, when the user asks whether this repo is ready for a swarm, or when a run failed and the cause might be the environment rather than the work. Re

• arbeitsplan-run

  Executes a compiled workflow as a redundant swarm — N candidates over the same scope in their own git worktrees, judged blind, exactly one landed. Use after arbeitsplan-compile, or when the user asks to run the workflow, start the swarm, or build several versions and pick one. Halts and surfaces rat

• arbeitsplan-start

  Names which of four approved workflows fits a stated task, the plugins to install for it, and the permission mode to run it in — and says plainly when none of the four fits. Use when the user asks where to start, which workflow or plugin combination applies, whether to use plan mode or auto mode for

• arbeitsplan-status

  Read-only report on an arbeitsplan run that has already started — phases done, candidates built and refuted, referee verdicts, open doubts, halts with their reasons, dispatch budget used, whether a lock is open, and the single next command. Use when the user asks how the run is going, what the swarm

• arbeitsplan-waves

  Installs an approved multi-wave design into a project as project-owned files — a pinned interpreter, the plan, a state helper, a path and runner guard, and one agent file per agent type — then launches and resumes it from the primary checkout. Use when a restructuring needs several waves of parallel

Subagents (12)

• adjudicator

  Use this agent for the final plan-mode ADJUDICATE phase of an arbeitsplan run — it reads the whole round (contract, candidates, blind verdicts, synthesis) and decides whether the round established what the contract asked, owning the cannotEstablish list. Never overturns a referee verdict — a disagre

• candidate-builder

  Use this agent when one candidate of an arbeitsplan fan-out phase must be built in its own git worktree, under exactly one assigned angle, against the acceptance criteria in workflow.json. Every dispatch names exactly ONE angle and ONE worktree; a dispatch naming several is out of scope and only the

• candidate-referee

  Use this agent to judge ONE arbeitsplan candidate against the acceptance criteria, given its diff and nothing else — no builder rationale, no angle, no other candidate, no prior verdict. One candidate per dispatch, always; a dispatch carrying two candidates is out of scope and only the first is judg

• cleaner

  Use this agent to PROPOSE a removal set for an arbeitsplan run — dead files, orphaned fixtures, stale analysis/arbeitsplan run directories, leftover worktrees and branches — each path with evidence that nothing reaches it. Read-only by design, it proposes, and scripts/sweep_artifacts.py (dry-run by

• contract-author

  Use this agent to turn an arbeitsplan INVENTORY into the acceptance contract every later phase runs under — criteria with runnable checks, plus the cannotCheck list of what no check can decide — BEFORE any candidate exists. Never writes code, and is never dispatched once candidates exist, because au

• implementer

  Use this agent for a single-writer arbeitsplan phase that implements one scoped change against a fixed contract — one writer, one worktree, one diff returned as data. Refuses fan-out (N writers over one scope is best-of-n with candidate-builder) and refuses to widen writeScope. Hands its diff to the

• inventory-extractor

  Use this agent to extract every item of one kind from ONE source partition of an arbeitsplan read-only fan-out — the map half of map-reduce-disjoint — returning a complete list with stable ids and paths, or an explicit truncation marker. One partition per dispatch; never summarises, never judges wha

• pattern-researcher

  Use this agent to propose which agentic pattern fits a stated problem, grounded first in the frozen catalog at references/patterns.md and only then in outside sources. Returns proposals with citations; it never writes workflow.json and never edits the catalog. A pattern the catalog marks rejected is

• refactorer

  Use this agent for a single-writer arbeitsplan phase that restructures code WITHOUT changing behaviour — rename, extract, move, inline — proving equivalence by running the same checks before and after. Refuses any behaviour change and refuses to run inside a fan-out phase (N restructurings of one sc

• scope-prover

  Use this agent to derive the concrete write scope a phase needs and, for a partitioned fan-out, to prove the partitions cannot overlap. Returns a proof or a refusal — a partition whose scope cannot be derived is reported as unprovable, never assumed disjoint. Never dispatched to widen a scope mid-ru

• script-runner

  Use this agent to run ONE declared command for an arbeitsplan `script` phase or design-table script node and return its exit code, the tail of its stdout, and its stdout JSON copied verbatim. The command arrives in the prompt, already rendered; the PreToolUse guard allows exactly that command, once,

• synthesizer

  Use this agent for the single-writer step of select-then-synthesize in an arbeitsplan run — given the refereed winner's diff and the runners-up, it returns the diff to land, borrowing a runner-up hunk ONLY where the spec's borrowGate names an acceptance criterion that hunk beats the winner on. Works

Hooks (2)

• PreToolUse

  python3 -B "${CLAUDE_PLUGIN_ROOT}/hooks/arbeitsplan_guard.py"

• Stop

  python3 -B "${CLAUDE_PLUGIN_ROOT}/hooks/record_stop_guard.py"

Similar plugins