Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Plan Implement

ASecurity

Implement an approved implementation plan (a `*.plan.json`, RFC 0030) one derived step at a time — `plan:next` names the step and what it covers, you implement exactly that, `plan:verify` records the result, one commit per step. Use when the user says "implement the plan", "next step", "continue the plan", "work through the plan", or names a plan file with the intent of building it.

29 stars
0 votes
0 copies
0 views
Added 9/23/2026
developmentbashtestingcode-reviewgitapidatabase

Works with

api

Security Analysis

A100/100

Scanned 9/23/2026

Install to Claude Code

$npx -y skills add gurenjs/guren --skill plan-implement --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Plan Implement?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Plan Implement
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/gurenjs-plan-implement/badge)](https://www.skillsdirectory.com/skills/gurenjs-plan-implement)

More formats (shields.io, HTML) on the badges page.

Download with Pro
Files
SKILL.md
---
name: plan-implement
description: Implement an approved implementation plan (a `*.plan.json`, RFC 0030) one derived step at a time — `plan:next` names the step and what it covers, you implement exactly that, `plan:verify` records the result, one commit per step. Use when the user says "implement the plan", "next step", "continue the plan", "work through the plan", or names a plan file with the intent of building it.
---

# Plan Implementation Skill

You implement a plan the way Guren derives it: step by step, in task order,
with completion read from the code rather than reported by you.

> The API rules in `__RULES_DIR__/` (orm-models, controllers-http, routes-codegen,
> testing) carry the verified signatures for what each step writes.

## The loop

```bash
bunx guren plan:next docs/plans/<slug>/plan.json          # the next step and its context (--json for the structure)
# implement exactly that step
bunx guren plan:verify docs/plans/<slug>/plan.json --step <id>
git commit                                                  # one step, one commit; name the step id in the message
```

Repeat until `plan:next` reports that every step is verified, or that no step
can be returned because every step left is held.

`plan:next` prints one step: its elements, the acceptance behaviours it must
write or see pass, and the verify commands. It never prints the whole plan, and
it refuses a working tree with uncommitted changes that are not the marked
step's own: finish or discard them first.

`plan:next`, `plan:verify` and `plan:waive` refuse a plan whose current hash no
approval names: one edited after it was approved, or one nobody approved. The
Stop hook stalls the marked step on it. Approving is the person's call, like
closing: report the refusal and wait. Do not run `plan:approve` yourself, and do
not edit the plan to get past it.

For an approved plan it also holds a step whose context changed after
approval: an element the step owns or names that another commit moved, so the
plan does not describe the application there. It lists each held step
with the element, how the step depends on it and what the reference checks say
now, and returns the next step that does not depend on one. A held step is a
person's decision (undo the change, or edit the plan so each stale element
states what the application holds now, such as an `existing` action another
commit renamed, and approve the edit): report it, and do not edit the
application back or the plan to make it pass.

## What a step asks for

- **`scaffold`**: run the generators it names (`make:feature`, `make:module`);
  do not hand-write what they emit.
- **`tests`**: write the behaviours as tests whose titles carry the acceptance id
  literally, `[AC-comments-1] a signed-in user can comment on a post`, and leave
  them failing. The step verifies with `tests:fail`, so a test that already
  passes, or is skipped, fails the step.
- **`data`**: the schema, migration and model; verified by `db:migrate` and
  `typecheck`.
- **`http`**: validators, controllers, routes, resources and policies, until
  `guren check` passes and the step's behaviours pass. An action is complete only
  when its route is mounted and validates through the route contract.
- **`pages`**: the Inertia pages; verified by `typecheck` and `guren check`.

Implement only the elements the step lists. An element of a later step is that
step's work, and `plan:status` will read it as drifted from the plan if it lands
elsewhere. Do not edit the plan file: a change of design is a revision
(`plan:revise`, when available), not an edit.

## What `plan:verify` records

The verdict is `verified`, `failed`, `blocked` or `incomplete`, under
`.guren/plans/<slug>.state.json` (git-ignored). `blocked` is the environment's:
a script the app lacks, a database that is unreachable, a timeout. It is not a
failure of the implementation, and it is not yours to route around: say what is
blocked and stop.

## The Stop hook

While a step is marked, the `Stop` hook verifies it whenever you end a turn and
sends you back while it is not verified, up to three times. It gives up, and
says why, when something the step names went stale since approval, when the
step or one of its elements is blocked, when nothing changed
since the last continuation, or after the third continuation. The step is then
recorded as stalled and `plan:next` returns it again, with the reason.

## What a stall means

A stall is a decision for a person, and there are three answers: fix the
environment, edit the plan (and approve it), or accept an element incomplete with

```bash
bunx guren plan:waive docs/plans/<slug>/plan.json <element-id> --reason "<why>"
```

which writes the reason into the decision log beside the plan, committed with
it. A waived element is left out of the step's judgement, so the loop moves on.

A waiver only lifts an element. A behaviour that fails makes its `tests`
command fail, and the step stays `failed` whatever is waived, so a behaviour
the code will not satisfy is a revision rather than a waiver.

The waiver is the person's decision, never yours. Report the stall, say which
of the three you think it needs and why, and wait to be told. Do not run
`plan:waive` to get past a step, and do not work around a stall in the code.

## After a task's last step

When the last step of a task verifies, ask the `code-review` agent to read that
task's diff against the plan elements the task covers. Its findings are
advisory: a reviewer asked for gaps reports some whether or not they exist, so
weigh each one against the plan before acting on it.

## When every step is verified

Closing the plan is the person's call, like approving it. Show them what
closing would write:

```bash
bunx guren plan:close docs/plans/<slug>/plan.json --dry-run
```

and run it without `--dry-run` only when asked. It refuses while any element is
neither verified nor waived, and names them. It writes the plan's document under
`docs/plans/` and draft blocks between `<!-- guren:plan … -->` markers in
`docs/entities/<Entity>.md`; edit the text outside the markers freely, since a
second close replaces only what is inside them.

Attribution

gurenjsgurenjs
View sourceMore from gurenjs →
SSkills DirectorySkills Directory

Your tool, in front of Claude Code builders.

3 founder slots · $299/mo · GSC-verified traffic · sponsors can never buy grades.

See placements

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Your tool, in front of Claude Code builders.

3 founder slots · $299/mo · GSC-verified traffic · sponsors can never buy grades.

See placements

Related Skills

Browser Extension Developer

Use this skill when developing or maintaining browser extension code in the `browser/` directory, including Chrome/Firefox/Edge compatibility, content scripts, background scripts, or i18n updates.

284072 votes

Seo Optimizer

SEO optimization with keyword analysis, readability assessment, technical validation, content quality. Use for search rankings, blog posts, content audits, or encountering keyword density, readability scores, meta tags, schema markup errors.

2192 votes

Google Official Seo Guide

Official Google SEO guide covering search optimization, best practices, Search Console, crawling, indexing, and improving website search visibility based on official Google documentation

1862 votes

Tanstack Start

Build a full-stack TanStack Start app on Cloudflare Workers from scratch — SSR, file-based routing, server functions, D1+Drizzle, better-auth, Tailwind v4+shadcn/ui. Use whenever the user mentions TanStack Start, asks to scaffold a full-stack Cloudflare app with SSR, wants an SSR dashboard, or asks for a React 19 + Cloudflare Workers app with file-based routing and server functions — even if they don't name TanStack Start specifically. No template repo — Claude generates every file fresh per ...

9881 votes

Pentest

PTES-aligned adversarial security audit for backend, frontend, and mobile applications. Produces a CVSS-scored Hacker Report with verified PoCs and phased remediation.

5491 votes
View all in development →