ワンクリックで
megaplan
Run the canonical Megaplan planning pipeline from its Arnold plugin home.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Run the canonical Megaplan planning pipeline from its Arnold plugin home.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Launch one durable, mutation-authorized meta-fixer to repair a failed automated fixer and its missed backstop, retrigger ordinary repair, and prove the identified epic or session advances. Use when a watchdog, repair loop, meta-repair loop, or progress auditor ran or should have run but the real chain stayed stuck, repeated stale evidence, claimed false success, or failed to self-correct. Invoke with `--target "EPIC_OR_SESSION_TEXT"`.
Launch an external model as a subagent for a second opinion, adversarial review, or delegated work. Default pathway is an agentic DeepSeek / Kimi / Zhipu GLM hermes subagent (file/web/terminal tools, one process or fanned out N-wide); also Codex (GPT-5.5) and Claude via the Agent tool. Use for independent root-cause analysis, cross-checking your reasoning, judge/jury panels, or handing implementation to a different model.
Scaffold a new Arnold native-first pipeline package from the _template skeleton.
Adversarial review of prose drafts by N reviewers, then revise. Not for code.
Survey every place loose work hides in a git repo — starting with the current checkout's own uncommitted/untracked/unpushed work, then local branches, all worktrees (`.megaplan-worktrees`, `.megaplan/bakeoffs/*`, agent-tool worktrees), stashes, detached HEADs, interrupted rebases/merges, submodules, remote branches on origin and other remotes, fork PR refs, sibling repo variants, other clones of this repo elsewhere on disk, the megaplan-cloud machine (Hetzner box) and every workspace on it, and GitHub Codespaces — classify each as land-on-main / delete / parked with reasoning, and act only on explicit per-item approval. Use when the user says "clean up loose branches", "prune branches", "what branches can I delete", "clean up worktrees", "review my stashes", "what's lying around in this repo", or asks for branch / worktree / stash housekeeping.
Steps to create a new Arnold native-first pipeline package from the template.
| name | megaplan |
| description | Run the canonical Megaplan planning pipeline from its Arnold plugin home. |
Use this pipeline for high-rigor planning workflows that need prep, plan, critique, gate, revise, finalize, execute, review, and tiebreaker stages.
For epic-sized, migration, public-contract, or otherwise drift-sensitive work,
capture durable end-state intent with a North Star anchor: pass
--north-star PATH to megaplan init for a standalone plan, or declare
anchors.north_star in chain.yaml for epics.
North Star questions are explicitly authored — they are not extracted from
code, diffs, or execution traces. In v1 the pipeline reads them from a
NORTHSTAR.md file in the plan directory (standalone plans) and/or from
sprint-brief anchors declared in chain.yaml (epic chains). There is no
semantic-parity extraction step: if a question is not written in one of those
sources it does not exist from the pipeline's point of view.
Gate (structured carry). When a gate or reviewer evaluator raises a
concern tied to a North Star question, it emits a north_star_actions[]
entry — an action with a concrete action_type (change_plan,
add_gate, add_scenario, add_checker, dead_delete, or
add_human_halt), a severity (blocking/advisory), supporting evidence,
and optional plan references. Actions in the dangerous categories
(route_authority, baselines, row_carrier_exemptions,
target_narrowing, generated_conformance_authority,
live_plan_topology_resume_risk) are blocking by schema rule regardless
of what the evaluator labels them. All actions are normalized at gate time
and carried forward through gate_carry.json (with gate.json as a
fallback).
Revise (change the plan or halt). The revise worker prompt receives
the carried north_star_actions[] with per-action-type instructions:
change_plan — make a concrete, traceable plan change.add_gate — add an explicit gate requirement.add_scenario — add a concrete scenario / test case.add_checker — add an automated checker.dead_delete — remove dead plan steps.add_human_halt — halt; the action cannot be mapped to a plan change.Before the revise worker runs, a pre-worker guard halts through the
existing CliError/step-failure path when any carried blocking action is
add_human_halt, has an unrecognized action_type, or is a
dangerous-category blocker with no concrete target (plan_refs or
required_change). Revise must change the plan or halt — it cannot
silently ignore a blocking action.
After the revise worker completes, the pipeline validates that every
carried blocking action has a corresponding north_star_actions_addressed[]
entry with concrete plan_refs and the matching action_type marker.
Prose-only resolutions, mismatched action types, and omitted actions are
rejected (fail-closed).
Finalize / Review → done (closeout blocking). Finalize refuses to
produce executable tasks if any carried blocking North Star action is
unresolved in the latest revise metadata. The review→done transition
runs a separate pre-check before TransitionPolicy.evaluate_review_done();
unresolved blocking actions deny the transition through the existing
denial path. Absent or malformed north_star_actions_addressed[] is
treated as all-carried-blocking-unresolved (fail-closed).
north_star_critical and robustnessChain and milestone specs support an optional north_star_critical: bool
(default false). When true, the milestone (or chain) declares that
North Star enforcement is mandatory for that work unit and cannot be
skipped.
north_star_critical: true is incompatible with bare and light
robustness. The chain-spec validator rejects these combinations with a
CliError — it never silently upgrades the robustness level. Acceptable
robustness levels are full, thorough, and extreme. Milestone-level
robustness takes precedence over driver-level robustness when both are
present; north_star_critical is OR'd across the chain-level and
milestone-level settings.
Some North Star action types conceptually suggest packaged skills (e.g. a
dead_delete action could map to a dead-code-removal skill), but no such
skills ship with the pipeline today. These are packaging follow-up
items — the pipeline enforces the structured action contract regardless,
and any future skill packaging would consume the existing
north_star_actions[] / north_star_actions_addressed[] metadata without a
contract change.