More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Execute the Plan: Small Verifiable Steps to Done

    1

    Carry out a written plan alone, task by task, without drift: a ledger as memory, a failing test before every change, a ruling for every judgment call, and one cold review at the end.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Execute the Plan: Small Verifiable Steps to Done

    Execute the Plan: Small Verifiable Steps to Done

    Example session with this skill installed

    I have a four-task plan for a fictional habit-tracking app called Streakline: add a habit, record a daily check-in, show the current streak, and send a weekly summary. The spec says streaks reset at midnight UTC and weekly summaries are opt-in. Run the plan with this skill and show me the ledger, one ruling, and the review outcome.

    • Read your context and instructions
    • Compiled the execute the plan

    SETUP — Plan and spec read end to end; suite green on a clean tree; ledger opened with the plan name, spec, and one-commit-per-task strategy.

    TASK 1, add a habit: test first (fails for the right reason), smallest change, plan verifications read in full, committed, ledgered. TASK 2, daily check-in: first run fault — the plan named no definition of done for this task, and the run proceeded without writing one. Caught at the task gate, fixed by writing the done-criteria into the ledger before starting: a check-in stores one entry per habit per UTC day, and a second check-in the same day is a no-op. TASK 3, current streak: ruling recorded — plan said streaks reset at local midnight, spec says UTC; spec wins. Ruling: UTC reset adopted, costs one timezone conversion at the boundary, worth it. TASK 4, weekly summary: implemented as opt-in per the spec, default off, with the opt-in flag stored per household rather than globally.

    FINISH — Cold review of the full diff found one important issue: the summary counted check-ins the streak display excluded, because the summary used local days. One fix pass aligned both on UTC; suite green after. Two naming nits went in the ledger. Branch diff is exactly the plan's four tasks plus one ruling.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    A finished plan deserves execution that matches it, and working alone takes away the two things a team gives you for free: a second pair of eyes per task and a memory that is not your own. This skill buys both back cheaply. Before task one you read the plan and the spec end to end, confirm the suite is green on a clean tree, and open a ledger that records every task, test result, ruling, surprise, and deferred item. Each task then runs the same loop: re-read the brief from the plan, write the test first and watch it fail for the right reason, make the smallest change that turns it green, run every verification the plan names and actually read the output, compare against the expected outcome, and ledger the result. Gaps in the plan get rulings, not stalls: the spec outranks the plan, the plan argues from the spec, and your judgment covers the rest, with every ruling naming what was chosen, why, and what it costs. Four stop conditions — destructive or irreversible actions, security-sensitive work, side effects outside the worktree, and a plan whose picture of the problem is wrong — are the only things that halt execution. The finish is a cold review of the whole diff plus one fix pass for critical and important findings, with minor findings ledgered instead of polished endlessly.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.

    • It cannot rescue a plan whose picture of the problem is wrong; that needs replanning, not execution.
    • It assumes a runnable test suite; without one, the per-task gate is weaker and the ledger must say so.
    • The cold review is one review at the end, not a fresh pair of eyes per task, so per-task correctness still rests on the tests.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 3 days ago

    Permissions

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    The spec wins, and the disagreement goes in the ledger as a ruling naming what was chosen, why, and what it costs. The plan is the spec's argument, not its replacement.

    Yes, and it must fail first for the right reason. A test that passes before the change proves nothing, and a test failing for the wrong reason guards the wrong thing.

    Only for the four stop conditions: destructive or irreversible actions, security-sensitive work, side effects outside the worktree, or a plan whose picture of the problem is wrong.

    New in Code Quality & Review