create-spec-plan
Create a detailed implementation plan based on an approved specification.
Skill by askidmobile
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/askidmobile-create-spec-plan
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/create-spec-plan
What's inside
- Skill name
create-spec-plan- Description
- Create a detailed implementation plan based on an approved specification. Analyzes the spec and the codebase, designs the architecture, breaks work into phases with checklists. Describes HOW we do it.
- License
- MIT
- Instructions
- 343 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Implementation Plan Generator (PLAN) You are an architect and tech lead. Your task is to create a detailed implementation plan based on a specification: **$ARGUMENTS** > **The plan describes HOW we do it: architecture, files, phases, checklists.** > Input is an approved specification from `docs/specs/`. ## PHASE 1: Load and context ### 1.1. Project recon (SILENT) Determine the context: 1. **Stack** — package.json, Cargo.toml, go.mod, requirements.txt, pyproject.toml, etc. 2. **Structure** — `ls` the root, main directories 3. **Existing plans** — check `docs/plans/`, `plans/`, `docs/` 4. **Task tracker** — TASKS.md, TODO.md, or equivalents. If a task-tracker script exists — use it to check for duplicates ### 1.2. Find the specification If `$ARGUMENTS` is a file path → read it. If `$ARGUMENTS` is a name → find it: ``` Glob: docs/specs/*$ARGUMENTS* ``` If not found → show the list: `ls docs/specs/` and suggest `/create-spec <name>`. ### 1.3. Spec analysis Read the spec in full. Extract: - All functional requirements (FR-XXX) - User scenarios and acceptance criteria - Non-functional requirements - Data model - Scope and constraints - **Deferred questions** (section "Deferred questions") — those whose `When the answer is needed` is "before plan" or "before plan phase N" must be closed in Phase 1.4. - **Spec decisions** — read as context (these are already-made architectural choices we build on). ### 1.4. Closing inherited questions from the spec If the spec has deferred questions that block the plan: 1. Write them into a queue. 2. Ask **one at a time** via `AskUserQuestion` — see [interactive questions rules](#interactive-questions-rules-shared-block). 3. After each answer **update the spec file**: - Move the entry from "Deferred questions" to "Spec decisions" (`D-NNN | Q | A | YYYY-MM-DD`). 4. Don't advance to Phase 2 until all blocking ones are closed. Also check the **brief** (`docs/brief/` folder: `open-questions.md` and `decisions.md`; or legacy single-file `docs/PROJECT-BRIEF.md`) — it may contain deferred questions that block the plan. Close them by the same protocol (update the brief's files). The brief's "Principles" section is binding for the plan's architecture choices. ## PHASE 2: Deep codebase analysis (SILENT) Based on the spec's requirements: ### 2.1. Affected files Find ALL modules, components, services, types, tests that will be touched. ### 2.2. Project patterns Study analogous implementations: - How similar modules are organized (structure, error handling) - How layering is organized (API → service → storage) - How data migrations work - How tests are organized ### 2.3. Dependencies - Which packages/libraries will be needed - Whether the required dependencies already exist ## PHASE 2.5: Surfacing and closing technical plan questions > **Goal**: the spec describes WHAT, the plan describes HOW. New forks appear at the boundary. Close them before generating the template so "Open questions" in the plan don't turn into an unresolved checklist. ### 2.5.1. Build the queue (SILENT) Typical sources for a plan: - **Architectural forks**: extend an existing module (coupling) vs create a new one (duplication); event-driven vs sync calls; in-process vs separate service. - **Data migration strategy**: in-place ALTER vs new table + bulk copy; blocking vs background. - **Rollback strategy**: feature flag vs git revert; data → can it be rolled back? - **Phase breakdown**: one big or 4 small; parallelizable vs strictly sequential. - **Test strategy**: what is unit-tested, what is integration-tested, what is manual; do we need golden / snapshot. - **Dependencies**: add a new package (supply-chain risk) vs write it ourselves. - **Edge cases not covered by the spec** that surfaced while reading the code. Empty queue → skip this phase, go to Phase 3. ### 2.5.2. Ask one at a time via AskUserQuestion Follow the [interactive questions rules](#interactive-questions-rules-shared-block). - One call = one question. - 2-4 options + auto Other. - Recommendation — first option with `(Recommended)`, if there is one. ### 2.5.3. What we record in the plan - **Closed questions** → into the new "Plan decisions" template section (Q→A with date). - **Deferred** → "Deferred questions" section with rationale. - **"Open without rationale"** must not exist in an approved plan. ## PHASE 2.7: Principles gate (SILENT unless violated) If the brief defines Principles (`docs/brief/PROJECT-BRIEF.md`): check every planned architecture choice against them. A violation is either redesigned away or explicitly justified in the plan's "Complexity & principle deviations" section — never silent. If the justification itself needs a user decision, add it to the Phase 2.5 queue. ## Interactive questions rules (shared block) When closing questions via `AskUserQuestion`, follow: - **One question = one tool call** (`questions` array of length 1). - **2-4 options** + automatic "Other".
From askidmobile/ai-spec-kit, MIT.
More from askidmobile/ai-spec-kit
- 0.525score
task-tracker
askidmobile
Manage project tasks in TASKS.md.
SkillBFreshness: BClaude Code +9 - 0.521score
wiki-compiler
askidmobile
Core compilation algorithm for the LLM Wiki Compiler.
SkillBFreshness: BClaude Code +9 - 0.521score
create-spec
askidmobile
Create a technical specification for a new feature or task.
SkillBFreshness: BClaude Code +9 - 0.521score
create-spec-implement
askidmobile
Phase-by-phase plan implementation.
SkillBFreshness: BClaude Code +9 - 0.521score
create-spec-review
askidmobile
Post-implementation retrospective.
SkillBFreshness: BClaude Code +9 - 0.520score
project-brief
askidmobile
Coverage-driven project brief engine.
SkillBFreshness: BClaude Code +9 - 0.518score
wiki-visualize
askidmobile
Launch a local interactive knowledge-graph visualization of the compiled wiki.
SkillBFreshness: BClaude Code +9 - 0.518score
wiki-init
askidmobile
Interactive setup of .wiki-compiler.json — mode, sources, output, article sections.
SkillBFreshness: BClaude Code +9 - 0.518score
wiki-migrate
askidmobile
One-time migration — replace raw file reads in the CLAUDE.md/AGENTS.md startup sequence with wiki articles.
SkillBFreshness: BClaude Code +9 - 0.518score
wiki-query
askidmobile
Answer a question from the compiled wiki instead of scanning raw source files.
SkillBFreshness: BClaude Code +9 - 0.518score
wiki-upgrade
askidmobile
Update the installed ai-spec-kit — git pull for symlink installs, reinstall for copies.
SkillBFreshness: BClaude Code +9 - 0.518score
commit
askidmobile
Commit and push in one go — reviews status and diff, writes a conventional commit message in the repo's style.
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Aug 9, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/askidmobile/ai-spec-kit
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from askidmobile
- 0.525score
task-tracker
askidmobile
Manage project tasks in TASKS.md.
SkillBFreshness: BClaude Code +9 - 0.521score
wiki-compiler
askidmobile
Core compilation algorithm for the LLM Wiki Compiler.
SkillBFreshness: BClaude Code +9 - 0.521score
create-spec-implement
askidmobile
Phase-by-phase plan implementation.
SkillBFreshness: BClaude Code +9 - 0.521score
create-spec-review
askidmobile
Post-implementation retrospective.
SkillBFreshness: BClaude Code +9 - 0.521score
create-spec
askidmobile
Create a technical specification for a new feature or task.
SkillBFreshness: BClaude Code +9 - 0.520score
project-brief
askidmobile
Coverage-driven project brief engine.
SkillBFreshness: BClaude Code +9
Cite this page
create-spec-plan skill. Agent Mart. https://agentmart.dev/skills/askidmobile-create-spec-plan (last updated 2026-10-24)