plan

スター0

フォーク0

更新日2026年7月20日 19:33

Create an execution-ready Parallax plan grounded in repository evidence and explicit verification.

インストール

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。

ソース

Master0fFate

Master0fFate/parallax-claudecode

ダウンロード

Plan

Pass the current Claude session ID to parallax_plan, then plan $ARGUMENTS without modifying files.

Read project instructions, manifests, architecture, tests, and the nearest comparable implementation.
Resolve material ambiguity. State the four invariants with concrete owners and paths.
Analyze nominal behavior, boundaries, errors, security, concurrency, compatibility, rollback, and observability.
For cross-subsystem or high-risk work, run the Hyperplan skill and incorporate only critiques supported by evidence.
Produce ordered, atomic items. Each item must name its files or component, observable acceptance criterion, and targeted check.
Identify assumptions, external blockers, and deferred work explicitly.

A good plan is short enough to execute, detailed enough for another agent to verify, and does not claim certainty unsupported by repository evidence.

このリポジトリの他の Skills

同じリポジトリ

Run a durable Horizon workflow for a multi-feature goal with bounded autonomous retries and an audit trail.

Compatibility entry point for the Parallax workflow when Claude resolves it as a plugin skill.

Report the current Claude session's Parallax gate, retries, verification, and trace status.

Implement a planned change under the Parallax write gate with friction-loop verification.

Record a Parallax protocol checkpoint with concrete evidence before gated implementation work.

Perform an evidence-based Parallax diagnosis or post-build audit and verify the repair.

name	plan
description	Create an execution-ready Parallax plan grounded in repository evidence and explicit verification.
argument-hint	[goal]
user-invocable	true

Pass the current Claude session ID to parallax_plan, then plan $ARGUMENTS without modifying files.

Read project instructions, manifests, architecture, tests, and the nearest comparable implementation.
Resolve material ambiguity. State the four invariants with concrete owners and paths.
Analyze nominal behavior, boundaries, errors, security, concurrency, compatibility, rollback, and observability.
For cross-subsystem or high-risk work, run the Hyperplan skill and incorporate only critiques supported by evidence.
Produce ordered, atomic items. Each item must name its files or component, observable acceptance criterion, and targeted check.
Identify assumptions, external blockers, and deferred work explicitly.

A good plan is short enough to execute, detailed enough for another agent to verify, and does not claim certainty unsupported by repository evidence.