Skip to content
Back to skills

Save Plan

ASecurity

Use when the user wants to save, export, or persist a plan (or any diagram-rich design doc shown in the CLI) to a markdown file so it can be opened and viewed with rendered mermaid diagrams. The terminal can't render mermaid; this writes the plan to docs/work/<slug>/plans/ (or a temp scratch file) for viewing in an IDE preview or on GitHub.

  • 3 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 5, 2026
ai-agentsgonodegitfrontendbackend

Works with

  • claude code
  • terminal
  • cli

Security analysis

A100/100

Scanned October 2, 2026

npx -y skills add addit-digital/addit-harness --skill save-plan --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Save Plan?

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

Security grade badge for Save Plan
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/addit-digital-save-plan/badge)](https://www.skillsdirectory.com/skills/addit-digital-save-plan)

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: save-plan
description: Use when the user wants to save, export, or persist a plan (or any diagram-rich design doc shown in the CLI) to a markdown file so it can be opened and viewed with rendered mermaid diagrams. The terminal can't render mermaid; this writes the plan to docs/work/<slug>/plans/ (or a temp scratch file) for viewing in an IDE preview or on GitHub.
user-invocable: true
argument-hint: "[short plan title] [--temp]"
---

# Save plan — persist a plan to a viewable markdown file

The Claude Code CLI shows a ` ```mermaid ` block as raw code, never a rendered
diagram, and a plan presented in the CLI isn't saved anywhere convenient to open.
This skill writes the plan to a markdown file so it can be opened with **rendered
mermaid** — in an IDE preview (VS Code, Ctrl/Cmd+Shift+V) or on GitHub
(renders mermaid natively).

## What to save
The current plan under discussion — typically the approved plan-mode plan (the
harness plan file) or the most recent plan/design doc written in this session.
Save it as-is when within its tier ceiling in references/doc-protocol.md; if over, trim to the plan template first and say what you cut. Keep mermaid blocks intact. If no
plan exists yet, say so and stop — don't invent one.

## Where it goes — artifact routing

This skill is for **implementation plans** (phased steps, to-do lists, acceptance
criteria). Docs are grouped by unit of work, not by type: one folder per feature,
bug fix, improvement, or enhancement at `docs/work/<slug>/`, holding everything
produced for it in per-type subfolders. The broader taxonomy:

| Artifact | Subfolder | Produced by |
|----------|--------|-------------|
| Implementation plan | `plans/` | This skill (default) |
| Architecture design / solution doc (no to-do list) | `solutions/` | `@backend-architect`, `@frontend-architect` directly |
| Architecture review report | `architecture-reports/` | `@architect-reviewer` directly |
| QA/regression verification report | `qa-reports/` | `@qa-engineer` directly |
| Per-change legal analysis | `legal/` | `@saas-legal-advisor` directly |

Layout, slug and collision: plugin conventions win — see rules/engineering-loop.md "Doc taxonomy".

For a pure architecture design, review report, or QA report, skip this skill —
the relevant agent writes directly to the correct subfolder inside
`docs/work/<slug>/`, following the same naming convention (create the folder + a
`docs/work/README.md` index row if missing).

## Choosing permanent vs temporary
- **Permanent** (default): `docs/work/<YYYY-MM-DD>-<kebab-title>/plans/plan.md`
  (or `docs/work/<ticket-id>-<kebab-title>/plans/plan.md` if the project uses an
  issue tracker), tracked in git. Best for design docs worth keeping and for
  remote sessions (push → view on GitHub).
- **Temporary** (`--temp`): a scratch file under `.plans/` (gitignored) for quick
  local viewing without committing. Mainly useful in **local** CLI sessions — in a
  remote/container session a temp file isn't on your machine, so prefer permanent
  + push there.

## Procedure
1. **Resolve the slug.** Use the given title (kebab-case it); otherwise derive a
   short slug from the plan's heading. Prefix with the issue-tracker ID if the
   plan is tied to a ticket, else `$(date +%F)-`. Then check
   `ls -d docs/work/<slug>`: if it exists and this session did not create or
   receive that slug, append `-2`, `-3`… until absent. Never overwrite another
   work item's plan.
2. **Pick the path.**
   - Default: ensure `docs/work/<slug>/plans/` exists (create it, and
     `docs/work/README.md` with an index if missing). Target
     `docs/work/<slug>/plans/plan.md` — the file is always named `plan.md`; the
     slug lives in the folder name, not the filename, so everything produced for
     this work item sits together.
   - `--temp`: ensure `.plans/` exists and is in `.gitignore` (add the line if
     missing); target `.plans/<slug>.md`. If not in a repo, fall back
     to the system temp dir.
3. **Write** the plan content to the file as prepared above (mermaid intact). Don't strip
   or "fix" diagrams.
4. **Update the index** (permanent only): add a row to `docs/work/README.md` —
   row per doc-protocol.md "Index row"; update in place.
5. **Report** the absolute path and how to view it rendered:
   - Local: "open in VS Code (or any markdown preview that renders mermaid) and toggle preview (Ctrl/Cmd+Shift+V)."
   - Remote/container session: "this file is in the container — commit & push,
     then view on GitHub (renders mermaid), or pull it locally."
   Don't commit or push unless the user asks.

## Notes
- Quality of rendering depends on valid mermaid syntax — keep node labels simple
  and quote labels containing special characters.
- This skill only persists and locates the file; it does not render images itself
  (no mermaid-cli/Chromium dependency). Rendering happens in the IDE/GitHub.

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.