Skip to content
Back to skills

Workspace Plans

ASecurity

Use this skill when writing, revising, or executing a workspace plan — the plan/README.md + phase-N-<slug>.md format the dashboard ingests. Don't use it for ad-hoc todo lists or in-repo design docs.

  • 7 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 5, 2026
ai-agentspythonnodeapi

Works with

  • api

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned September 25, 2026

npx -y skills add atretyak1985/swarmery --skill workspace-plans --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Workspace Plans?

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

Security grade badge for Workspace Plans
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/atretyak1985-workspace-plans/badge)](https://www.skillsdirectory.com/skills/atretyak1985-workspace-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: workspace-plans
version: "1.0.0"
owner: "swarmery-core"
description: "Use this skill when writing, revising, or executing a workspace plan — the plan/README.md + phase-N-<slug>.md format the dashboard ingests. Don't use it for ad-hoc todo lists or in-repo design docs."
color: cyan
docs:
  status: draft
  updated: 2026-09-01
---

# Purpose

The private workspace plan format is a contract: the control plane parses it
to render plan progress, phase summaries, and spec coverage. A plan that
deviates from the format ships work the operator cannot see.

# The contract (never violate)

- Plans live at `<workspace>/<project>/workspace/working/{YYYY}/{MM}/{DD}/{slug}/plan/`;
  task-id `yyyy-mm-dd-slug` is derived from the path — never encode the date
  in the folder name.
- `plan/README.md` — objective, architecture decisions with real file paths,
  the sequencing table with exactly these header cells:
  `| # | Phase | Doc | Depends on |`, risks, Definition of Done.
- `plan/phase-N-<slug>.md` per phase (flat; `step-NN-*.md` is legacy
  read-compat only) — each with a self-contained copy-paste executor prompt
  whose Verify block is file-backed (a repo script or test by absolute path;
  no heredoc, `python3 -` or `node -e` — see `resources/plan-format.md`),
  `- [ ]` acceptance criteria checkable by command or boolean inspection, and
  an empty `## Completion Report` section as the LAST section.
- A `## Forecast` section in each phase doc, immediately before that stub,
  holding one `kind: prior` yaml block (see `resources/plan-format.md`).
  It is a prediction, not a limit: do whatever the phase actually needs.
- `plan/manifest.json` — machine-readable phase DAG (must pass
  `python3 -m json.tool`).
- Optional `plan/spec.md` with `- [ ] **SC-n** — <criterion>` lines; then
  every phase doc carries a `**Covers:** SC-…` header line (coverage is
  linted).

# Executor duties

After reading the code the phase touches and before your first edit, add a
`kind: posterior` block beside the prior in `## Forecast` — the
prediction-not-a-limit rule above applies unchanged — and if the work then
diverges from it, say how and why in a short "Where reality diverged"
paragraph of the Completion Report.

Tick each satisfied criterion `- [ ]` → `- [x]` immediately after verifying
it — progress is derived only from these checkboxes. When a phase's last
criterion is ticked, fill that doc's `## Completion Report` (what shipped,
files/commits, verification output, deviations, plus three mandatory fields —
**Blocked calls**, **Delegation cost**, and the one-sentence **What would have
made this cheaper**; ≤50 lines) — the dashboard renders exactly that heading.
Each of the three is written out even when empty ("none" / "Nothing, it was
already minimal"); `resources/plan-format.md` carries their exact wording and
is the source of truth. Never tick unsatisfied criteria; never archive
with unmet criteria or a missing SUMMARY.md.

Full field-by-field format, examples, and the manifest schema:
`resources/plan-format.md`.

# How to use

## What it does

Carries the exact workspace-plan format the dashboard ingests: README sequencing table, phase docs with file-backed-verification executor prompts and checkbox criteria, manifest DAG, optional spec with SC coverage, and the tick/Completion Report duties for executors — including the report's Blocked calls, Delegation cost and "what would have made this cheaper" fields.

## When to use it

Load it whenever you write a new plan, revise one, or execute one — any time plan files under a task dir's `plan/` are created or updated.

## How to invoke

The `@core:planner` and `@core:implementation-agent` agents list it in their `skills:`; load it manually with `/workspace-plans` context or by reading `resources/plan-format.md` when editing plan files by hand.

## Worked example

A planner writing `plan/phase-2-api.md` includes: `**Covers:** SC-3`, a prompt an executor can run cold whose Verify block names `scripts/check-api.sh` rather than inlining a heredoc, three `- [ ]` criteria each with a check command, and ends the doc with an empty `## Completion Report` section. The executor later ticks the boxes one by one and fills the report — Blocked calls, Delegation cost and the one-sentence cheaper field included — and the dashboard's phase Summary tab shows exactly that section.

Files in this skill

  • SKILL.md3 KB
  • resources/plan-format.md3.8 KB

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.