Skip to content
Agentmart
Submit a tool

plan-mode

Create and maintain implementation plans only.

Skill by karthrand

IndexedMIT
Write a review

Chalk Labs

Need help deploying plan-mode?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

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/karthrand-plan-mode

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.

We use your email only for this newsletter. Privacy

About

Create and maintain implementation plans only. Use when the user wants requirement clarification, codebase-aware planning, technical research, phased breakdown, plan updates, or sub-plans before any coding. Read the project, verify external technical facts, and write plan documents under the project root `.plan/` directory. Do not execute plans or modify source code.

Use cases: writing, research

No 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/plan-mode

What's inside

Skill name
plan-mode
Description
Create and maintain implementation plans only. Use when the user wants requirement clarification, codebase-aware planning, technical research, phased breakdown, plan updates, or sub-plans before any coding. Read the project, verify external technical facts, and write plan documents under the project root `.plan/` directory. Do not execute plans or modify source code.
License
MIT
Instructions
123 lines in SKILL.md · view on GitHub

Files (10)

  • README.md
  • SKILL.md
  • openai.yaml · agents/
  • example-index.json · references/
  • example-parent-plan.md · references/
  • example-sub-plan.md · references/
  • examples.md · references/
  • plan-template.md · references/
  • research-policy.md · references/
  • sub-plan-template.md · references/

SKILL.md (first lines)

# Plan Mode

## Overview

Use this skill to produce planning artifacts before implementation.
Inspect the project, clarify scope, verify technical facts, and write plans under `.plan/`.

## Enforce Boundaries

- Plan only. Do not implement, refactor, migrate, or edit source code.
- Write only under `{project-root}/.plan/`.
- Do not invoke execution workflows, `do-plan`, task-sync systems, or platform-specific orchestration.
- Mark any unverified statement as `待确认`.
- Ask at most one focused user question at a time.
- Default plan titles, filenames, and section titles to Chinese.

## Detect the Project Root

Resolve the root in this order:

1. Use the Git repository root.
2. Otherwise use the nearest directory containing a project marker such as `package.json`, `pyproject.toml`, `Cargo.toml`, `go.mod`, or `.git`.
3. Otherwise use the current working directory.

Create and maintain this structure:

```text
{project-root}/.plan/
├── current/
├── archive/
└── index.json
```

## Follow the Workflow

1. Classify the request as:
   - create a new plan
   - update an existing plan
   - split a step into a sub-plan
2. Inspect the local project before asking the user about codebase facts.
3. Ask one focused clarification question only when local inspection cannot resolve a material ambiguity.
4. Read `references/research-policy.md` before doing external research.
5. Draft the parent plan with `references/plan-template.md`.
6. Split oversized or high-risk steps with `references/sub-plan-template.md`.
7. Write the results to `.plan/current/`.
8. Update `.plan/index.json` with lightweight metadata.
9. Return a concise summary and stop for user review.

## Apply Research Rules

Use this source priority:

1. Context7
2. Existing search tools in the current environment
3. General web search

Apply these rules:

- Prefer primary sources such as official documentation, standards, maintained repositories, release notes, and maintainer statements.
- State when Context7 is insufficient before falling back to search tools.
- Attach a source link to every external factual claim in the plan.
- Label conclusions derived from sources plus local code inspection as `推断`.
- If sources conflict, state the conflict and prefer the newer or more authoritative source.

## Write Plan Documents

Use these rules for every parent plan:

- Save parent plans in `.plan/current/`.
- Use filenames in the format `YYYY-MM-DD-中文计划名.md`.
- Keep a machine-friendly `plan_id` in metadata.
- Separate `已确认信息` from `待确认问题`.
- Give every step:
  - 目标
  - 涉及文件或模块
  - 具体操作
  - 完成标志
  - 验证方法
- Add a mitigation for every high-risk item.
- Add explicit acceptance criteria and verification steps.

Move completed or abandoned plans to `.plan/archive/` without changing `plan_id`.

## Create Sub-Plans Deliberately

Create a sub-plan during planning, not execution, when any of these is true:

- One step touches more than 3 files.
- One step contains multiple independent milestones.
- One step needs a separate tradeoff analysis.
- One step carries high risk or a long verification chain.
- One step remains unclear after a short breakdown.

Apply these rules:

- Limit the total depth to 2 levels.
- Place sub-plans inside a folder named after the parent plan filename without `.md`.
- Use filenames such as `01-接口梳理子计划.md`.
- Link every sub-plan back to the parent plan and parent step.
- Keep only summary-level intent in the parent plan for any step that has a sub-plan.

## Maintain the Index

Treat `.plan/index.json` as metadata, not source of truth.

Include these fields in each record:

- `plan_id`
- `title`
- `path`
- `status`
- `updated_at`
- `parent_plan_id`
- `children`

## Load References As Needed

- Use `references/plan-template.md` for parent plan structure.
- Use `references/sub-plan-template.md` for sub-plan structure.

From karthrand/karthrand-ai-public, MIT.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Mar 15, 2026 (7 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from karthrand

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get plan-mode running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying plan-mode?

CallGet help
Cite this page

plan-mode skill. Agent Mart. https://agentmart.dev/skills/karthrand-plan-mode (last updated 2026-10-24)