Skip to content
Back to skills

Executing Plans

ASecurity

Use when you have a written implementation plan to execute inline in this session, with checkpoints - the non-fleet execution path.

  • 2 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 3, 2026
ai-agentsgonode

Works with

  • cli

Security analysis

A100/100

Scanned September 23, 2026

npx -y skills add andreymudri/claude-teammates --skill executing-plans --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Executing Plans?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Executing Plans
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/andreymudri-executing-plans/badge)](https://www.skillsdirectory.com/skills/andreymudri-executing-plans)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: executing-plans
description: Use when you have a written implementation plan to execute inline in this session, with checkpoints - the non-fleet execution path.
---

# Executing Plans

_Adapted from the MIT-licensed superpowers plugin by Jesse Vincent. See NOTICE.md._

Execute a written plan inline, in this session, task by task, checkpointing with your human
partner as you go. This is the counterpart to `parallel-execution`: same run state, different
cadence. Use `using-fleetmates` to decide which one applies before reaching for this skill.

## 1. Initialize the run

    node "<fleetmates root>/scripts/cli.mjs" init-run <planPath> --run <runId> --root <project root>

This writes `.fleetmates/<runId>/plan.json` and `status.json` — the same shared state a fleet
run would write. Because of that:

- `node "<fleetmates root>/scripts/cli.mjs" digest --run <runId> --root <project root>`
  renders this run unchanged, whether it ran inline or across teammates.
- `fleet-lifecycle resume <runId>` can resume it later.
- If the work turns out to be bigger than expected mid-run, hand it to `parallel-execution`
  without translating anything — the state is already in the shape it expects.

Never start implementation on `main`/`master` without explicit consent from your human partner.

## 2. Review the plan critically

Read it fully before touching code. Raise concerns with your human partner before starting —
gaps, unclear steps, or a task that looks bigger than a single sitting.

## 3. Execute and checkpoint

Work through tasks in dependency order. Follow each step; run the verifications it specifies.

Checkpoint with your human partner **per task, or per small batch of related tasks** — not
automatically per phase the way a fleet gate does, and not silently at the end of the whole
plan. Report what changed, what you verified, and any deviation from the plan before moving on.

### Gate manifest is optional here

A fleet phase demands a `fleetmates.gate.json` manifest before it will pass. Inline work does
not need one. When no manifest exists, checkpoint against the plan's own verification steps
instead of demanding one be written first.

**Do not require a phase structure, a manifest, or a reviewer dispatch for small inline work.**
Ceremony that makes a two-task change expensive is ceremony that gets skipped — and a skipped
skill protects nothing. Reserve that machinery for work that has actually grown into a fleet
(see `parallel-execution` and `phase-gate`).

## 4. Finish

After all tasks are complete and verified, use `finishing-a-development-branch` (if available)
or otherwise confirm tests pass and present integration options to your human partner.

## When to stop and ask for help

Stop immediately when you hit a blocker (missing dependency, failing test, unclear
instruction), when the plan has a gap that prevents starting, or when a verification fails
repeatedly. Ask rather than guess.

## Remember

- Shared state with the fleet path: `.fleetmates/<run-id>/`, `init-run`, `digest`, resumable via
  `fleet-lifecycle`.
- Checkpoint per task or per batch, with the user — not per phase, not automated.
- No gate manifest required; the plan's own verification steps stand in for one.
- Scale up to `parallel-execution` any time without rewriting state.

Attribution

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

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.