Skip to content
Back to skills

Plan Operations

ASecurity

Execute approved plans continuously with truthful stage-end writeback.

  • 11 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 23, 2026
ai-agents

Security analysis

A100/100

Scanned October 6, 2026

npx -y skills add startmeupai/swe-agents --skill plan-operations --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Plan Operations?

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

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

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: plan-operations
description: Execute approved plans continuously with truthful stage-end writeback.
---

# Plan Operations

## Trigger Conditions

Use when asked to execute, continue, update, or finish an approved plan.

## Required Inputs

- Plan path, current state, stage owners, dependencies, and verification criteria.

## Workflow

1. Read the entire plan and confirm approval, status, scope, and blockers.
2. Check shared-checkout state and assign disjoint specialist ownership.
3. Execute the next unblocked stage without pre-filling later status.
4. Run the stage's named checks and preserve exact results.
5. Reconcile each checkbox independently against its own wording. The task
   boxes are the writeback; derive stage and plan status from them and never
   edit status on its own.
6. Tag every box left open at a stage boundary with why it is open:
   `**Browser gate:**`, `**Provider gate:**`, `**Deployment gate:**`,
   `**Operations gate:**`, `**Legal gate:**`, `**Human gate:**`, or
   `**Blocked on <owner or dependency>:**`.
7. Write status, evidence, blockers, and next dependency into the plan.
8. Continue until complete, genuinely blocked, or at a human-only gate.

## Stage Status

| Status | Meaning |
| --- | --- |
| Pending | Not started; open boxes need no tag |
| In progress | Work is underway inside the current run |
| Blocked | Stopped on a named owner or dependency; every open box is tagged |
| In review | Development is done; every open box is a tagged gate |
| Complete | Every box is checked |

## Deterministic Checks

- The project's plan checker (for example `pnpm check:plans`) exits 0 at every
  stage boundary: statuses are known, Complete stages have no open box, and
  every open box in a Blocked or In review stage carries a gate tag. Without a
  checker, verify the same conditions by reading the plan.

## Safety and Permission Boundaries

- Do not expand scope, publish, deploy, or perform destructive operations without authority.

## Required Evidence

- Stage-local command results, observations, changed files, and blocker ownership.

## Completion Condition

- All executable stages are complete and remaining external/human gates are explicit.

## Example

`Execute the approved Project Alpha plan and write back after every stage.`
See `examples/plans/stage-end-writeback.md` and
`examples/plans/review-gate-writeback.md` in the reference repository.

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…