القائمة
Creates executable implementation plans that break down a brainstorming design into detailed, unambiguous per-batch tasks for execution.
يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.
اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
يتجاوز الأمر المباشر Prompt المخصّص للمراجعة. افحص المصدر قبل تشغيله.
npx skills add https://github.com/FradSer/dotclaude --skill writing-plansيبقى الأمر في سطر واحد. مرّر أفقيًا لمراجعته كاملًا قبل النسخ.
تفضّل نسخة محلية؟ نزّل الملفات المتاحة حاليًا لدى SkillsMP.
Reviews a pull request: runs its own baseline review of the PR diff, then a persistent Monitor watches CI and incoming reviewer comments, triages each comment through an independent skeptical agent, applies only verified fixes, and commits+pushes via inline git commands until CI passes and no comments remain to adopt — then asks whether to merge. Use this skill when the user asks to "review a PR", "monitor PR review comments", "address reviewer feedback on #123", or "watch CI on a pull request".
Creates atomic conventional git commits using git-agent and pushes changes to the remote repository.
Creates atomic conventional git commits using git-agent. This skill should be used when the user requests "commit", "git commit", "create commit", or wants to commit changes using AI atomic commits.
عرض SKILL.md
استنادا إلى تصنيف SOC المهني
| name | writing-plans |
| description | Creates executable implementation plans that break down a brainstorming design into detailed, unambiguous per-batch tasks for execution. |
| argument-hint | ["design-folder-path"] |
| user-invocable | true |
| disable-model-invocation | true |
| allowed-tools | ["Read","Write","Edit","Glob","Grep","Agent","Bash(git-agent:*)","Bash(git:*)","Bash(${CLAUDE_PLUGIN_ROOT}/lib/seed-checklists.sh:*)","Bash(${CLAUDE_PLUGIN_ROOT}/lib/docs-index.sh:*)"] |
Create executable implementation plans that reduce ambiguity for whoever executes them — substantial Phase 1 → Phase 6 work, run wrapped in /goal (see below).
/goalLaunch it under Claude Code's built-in /goal (v2.1.139+) so the run continues to a committed plan instead of stopping mid-phase:
/goal "Claude has narrated a successful plan commit (with commit hash) and reported the Phase 4 reflection sub-agent verdicts inline" /superpowers:writing-plans <design-path>
/goal is a user-typed outer wrapper (a skill cannot enable it for itself mid-run), and its evaluator judges only what Claude narrates in the transcript — phrase the condition against narrated output (the commit hash, the literal Phase 4 reflection summary), never filesystem state. Full semantics and condition phrasing: ../../skills/references/goal-wrapper.md.
Read bdd-specs.md from the resolved design folder. Count Scenario: occurrences. Bail out — do NOT decompose tasks — when EITHER:
bdd-specs.md < 3, OR2× BDD scenarios + 1 setup task, or prefer an explicit "Task Estimate" hint in the design _index.md)The OR-gate catches the "2 BDD + many setup tasks" shape an AND-gate lets through.
Bail-out response (output verbatim):
Design too thin for full task-decomposition pipeline (BDD < 3 OR estimated tasks < 5). Drafting a one-page lightweight plan inline instead. To force the full pipeline, re-invoke as
/superpowers:writing-plans --force <design-path>.
Then write a single _index.md (no per-task files, no Phase 4 reflection, no plan evaluator) and exit. The --force token (literal in $ARGUMENTS, case-sensitive, whole-token match) bypasses this check.
Read _index.md from the resolved design folder. Bail out — do NOT decompose tasks — when grep matches any of:
grep -nE "STATUS:.*NOT.JUSTIFIED|DESIGN-NOT-YET-JUSTIFIED|DESIGN-CONSIDERED-DEFERRED|DO NOT IMPLEMENT" "<resolved-design-folder>/_index.md"
This catches designs explicitly marked "not approved to advance" (full semantics: docs/retros/checklists/design-v1.md JUST-01; inciting case: docs/retros/2026-05-09-v3-considered-deferred.md).
On match, refuse deterministically (the marker is dispositive — do not interpret it away):
Refusing: <design-path>/_index.md:{N} is marked NOT-JUSTIFIED — '{matched text}'. Re-invoke /superpowers:brainstorming to revise the design, or pass --justify-override to bypass this gate.Override: Pass --justify-override (literal token in $ARGUMENTS, case-sensitive, whole-token match) to bypass this refusal and continue to First Action.
PROHIBITED: Do NOT conflate --force (bypasses the size gate only) with --justify-override (bypasses this justification gate only) — they are independent; --force on a thin design does not bypass a NOT-JUSTIFIED refusal.
$ARGUMENTS provides a path (e.g., docs/plans/YYYY-MM-DD-topic-design/), use it*-design/ folder whose basename sorts last under YYYY-MM-DD-*-design/ — highest date prefix in the name, not filesystem mtime: ls -1d docs/plans/*-design/ 2>/dev/null | sort | tail -1 (NOT ls -t — edited old folders outrank fresh ones by mtime). Use the result directly; do NOT pause to confirm.*-design/ folder exists in docs/plans/, refuse with: Refusing: no design folder found under docs/plans/. Run /superpowers:brainstorming first, or pass the design folder path explicitly. Then exit._index.md and bdd-specs.md. Consult the docs index before drafting: run bash "${CLAUDE_PLUGIN_ROOT}/lib/docs-index.sh" show <design-path>. If the status is expired:, REFUSE — output a one-line note citing the expired status and exit without creating a plan folder. Then consult memory: run bash "${CLAUDE_PLUGIN_ROOT}/lib/docs-index.sh" list --kind memory --status active and Read the top topically-relevant matches before Phase 1.bdd-specs.md completely. This is the source of truth for your tasks.Core Concept: Explicit over implicit, granular tasks, verification-driven, context independence.
MANDATORY: Unit-test tasks require external dependency isolation with test doubles (DB/network/third-party APIs). All other decomposition invariants are enforced by Phase 2 steps 1-7.
Define goal, architecture, constraints, and context.
bdd-specs.md from the design folder../references/structure-template.md to outline the plan.## Context in _index.md: why this work is needed (motivation, constraints, prior incidents); for existing-code changes add a current-state vs target-state comparison table (omit for greenfield work).## Global Constraints section in _index.md — the cross-task invariants every batch and every task must respect (these make parallel batches safe). One invariant per bullet, no prose; omit only when the design carries zero cross-cutting constraints (rare). Categories and examples: ./references/structure-template.md §Global Constraints.Break into small tasks mapped to specific BDD scenarios.
PROHIBITED: Do not ask the user to choose granularity, approve decomposition, or confirm the plan mid-process — apply steps 1-7 plus these additions automatically; Phase 4 reflection is the quality gate and the post-commit git show diff is the audit surface.
NNN before feature pairsbdd-specs.md, so the executor sees the complete scenario without switching files. Format: ./references/structure-template.md §BDD Scenario.npm test tests/login.spec.ts).task-NNN-<feature>-test) must precede its paired impl task (task-NNN-<feature>-impl) via depends-on.**depends-on** field listing only true technical prerequisites — tasks whose output is required before this task can start. PROHIBITED: chaining tasks just to impose execution order. Independence rules and examples: ./references/task-granularity-and-verification.md §Dependency Rules..md file per task. Filename pattern: task-<NNN>-<feature>-<type>.md (e.g. task-001-setup.md, task-002-feature-test.md / task-002-feature-impl.md — test and impl for the same feature share the NNN prefix; types: test, impl, config, refactor; full template: ./references/structure-template.md §Task Files).
task-001-setup is justified only when 2+ downstream tasks share its output; pseudo-independent setup tasks create false parallelism and inflate batch counts.## Interfaces section (after ## BDD Scenario) declaring the contracts this task exposes or consumes (function signatures, types, API endpoints, event names, CLI flags). depends-on-linked tasks connect through these blocks, so parallel batches can verify interface compatibility before merge.## Interfaces block define contracts, but never the body logic.Verify completeness, save, advance to Phase 4 in the same iteration.
docs/plans/YYYY-MM-DD-<topic>-plan/ folder.
_index.md MUST include "Execution Plan" section with inline YAML metadata_index.md MUST include "Task File References" section linking all task files_index.md MUST include "BDD Coverage" section confirming all scenarios are covered_index.md MUST include "Dependency Chain" section with visual dependency graph (will be populated in Phase 4)./references/structure-template.md §Execution PlanBefore committing, verify plan quality with parallel fresh sub-agents, each in an isolated context (no pollution from the main planning session).
Launch three sub-agents in parallel via the Agent tool, subagent_type=general-purpose (full prompt templates: ./references/reflection.md):
Additional sub-agents (launch as needed): Red-Green Pairing Review, File Conflict Review.
Integrate and Update:
_index.md "Dependency Chain"git show; do NOT pause for approvalOutput: Updated plan, dependency graph in _index.md, reflection summary inline — ready for Phase 5.
Mid-stream cancellation (only under /goal; user injects "abort"/"cancel"/"start over"): stop with a one-line note — no commit, no Phase 5; the user re-invokes with the new framing if they want to retry.
Multi-turn resumption (only when wrapped in /goal after an interrupted turn): do not restart from Phase 1 — Glob "docs/plans/*-plan/_index.md", read the in-progress plan folder, and resume at the phase the filesystem indicates: task files exist but _index.md lacks the YAML tasks: block or References/Coverage sections → Phase 3; _index.md complete but no "Dependency Chain" graph → Phase 4; plan complete, no commit → Phase 5; already committed → Phase 6.
The sub-agents are the sole plan-quality reviewer (no separate plan-mode evaluator; the user reviews post-commit via git show). They read docs/retros/checklists/plan-v{N}.md as their rubric. On any checklist-item FAIL, fix the offending plan files and rerun that sub-agent — do NOT commit a plan with an unaddressed FAIL.
Auto-seed checklist when missing: before spawning the reflection sub-agents, if no plan-v{N}.md exists, run bash "${CLAUDE_PLUGIN_ROOT}/lib/seed-checklists.sh" plan docs/retros/checklists/plan-v1.md (exit 0 = seeded, 3 = already exists → proceed, 1/2 = abort), then pass the resolved checklist path to each sub-agent prompt.
MUST: each reflection sub-agent prompt MUST include an instruction to read the resolved checklist file and apply each item as a binary PASS/FAIL rubric (the template in ./references/reflection.md §Checklist Binding carries this directive — do not strip it). A sub-agent without the checklist produces an unanchored opinion, not a verdict.
Commit the plan folder using git-agent (with git fallback).
Actions:
0. Upsert the plan into the docs index (before git add): run bash "${CLAUDE_PLUGIN_ROOT}/lib/docs-index.sh" upsert plan <new-plan-path> --status active --summary "<one-line>". CRITICAL do-not-defer — the index update lands in the same commit-group as the plan folder.
0.5. CRITICAL do-not-defer: Conditional memory-write step, gated on a Phase 4 sub-agent FAIL that required a fix-and-rerun (not a first-pass PASS) — if triggered, write docs/memory/pitfall_<slug>.md capturing the false-positive cause, then run bash "${CLAUDE_PLUGIN_ROOT}/lib/docs-index.sh" upsert memory docs/memory/pitfall_<slug>.md --status active --summary "<one-line>" --category pitfall. No-op if every sub-agent passed first try.
git add is denied by the git plugin's hook: git add docs/plans/YYYY-MM-DD-<topic>-plan/ && git-agent commit --no-stage --intent "add implementation plan for <topic>" --co-author "Claude <Model> <Version> <noreply@anthropic.com>"--free flag/git:commit skill via the Skill tool; full ladder in ../../skills/references/git-commit.mdOutput:
Plan complete. To execute this plan, use
/superpowers:executing-plans.
PROHIBITED: Do NOT offer to start implementation directly.
Clear goal/constraints, decomposed tasks with file lists and verification, BDD steps, commit boundaries, no vague tasks, reflection complete with all FAILs addressed, git commit landed.
./references/structure-template.md - Plan and task file templates./references/task-granularity-and-verification.md - Task breakdown, verification, dependency rules./references/reflection.md - Sub-agent prompts for plan reflection../../skills/references/git-commit.md - Git commit patterns../../skills/references/goal-wrapper.md - /goal semantics and condition phrasing (shared)