Skip to content
shell
$ npx -y skills add aleksander-dytko/ai-pm-workspace --skill guide --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/guide
How auto-invocation works

Context preview

The summary Claude sees to decide when to auto-load this skill.

Interactive 8-module learning path for the shipped skills, role-adaptive

SKILL.md

guide.SKILL.md
name: guide
description: Interactive 8-module learning path for the shipped skills, role-adaptive
argument-hint: "[module number 1-8, or 'next', or 'reset']"

Guide

An interactive tour through the shipped skills. Each module teaches one skill by running it against sample data in `samples/`. Designed for PMs, designers, analysts, or engineers - the skill adapts framing to the role set in `CLAUDE.md`.

Pre-emit checklist (run BEFORE every chat message in this skill)

Before you send any message while `/guide` is active, audit your draft against these rules. If any fails, fix it before sending.

1. **Vault paths are backticked.** Any token that looks like a file path (contains `/` or ends in `.md`) must be wrapped in backticks so Claude Code renders it as a clickable reference. A path without backticks is a bug - it renders as plain text and the user can't click it. 2. **File-argument commands use `@`.** When you tell the user how to run a skill with a file input, the path must be `@`-prefixed so Claude Code's file autocomplete activates: `` `/meeting @samples/sample-meeting-transcript.md` ``, not `` `/meeting samples/sample-meeting-transcript.md` ``. 3. **Module intros below are printed verbatim.** Each module has a literal intro block in this file. Print that block exactly as written - do not rephrase, do not add extra paragraphs, do not strip backticks. The blocks are short on purpose. 4. **Gate order.** If the current turn is a first-run orientation, a resume, or a numeric jump, the message ends with `Ready to start Module N? (y/n/skip)` and nothing else - no module intro on the same turn. If the user just typed `next`, there is NO gate - go straight to the next module's verbatim intro.

Progress tracking

Progress is stored in `.claude/memory/guide-progress.md`. The `.claude/memory/` directory ships with the template (see its `README.md`), so do NOT run `mkdir` or `ls` to check - use the `Read` tool directly, and if Read returns "file does not exist", treat it as a first run and create the file with `Write`.

When you create the progress file for the first time, use this content (substitute today's date for `YYYY-MM-DD`):

---
tags: GuideProgress
started: YYYY-MM-DD
---

# /guide progress

- [ ] Module 1: /meeting
- [ ] Module 2: /today morning
- [ ] Module 3: /decision
- [ ] Module 4: /communicate
- [ ] Module 5: /create-epic
- [ ] Module 6: /competitive-research
- [ ] Module 7: Capstone
- [ ] Module 8: Extend your workspace

Mark a module complete (`[x]`) only once the user finishes it (see "End-of-module recap" for the specific moment).

Mode selection

Keep tool calls minimal and silent - no `Bash ls`, no `mkdir`, no "let me check..." narration. The user wants a coach, not a script.

Check `$ARGUMENTS`:

  • **Empty** (`/guide` with no args):
  • Read `.claude/memory/guide-progress.md`.
  • If Read returns "file does not exist", it's a first run: show the **First-run orientation block** (verbatim), create the progress file with `Write`, then stop. The orientation block ends with the "Ready to start Module 1? (y/n/skip)" prompt. On the user's next turn with `y`, print the **Module 1 verbatim intro** and stop.
  • If progress exists, show the **Resume line** (one line: `` You're on Module N of 8 - [Module title]. `/guide <n>` to jump, `/guide reset` to start over. ``), then ask `Ready to start Module N? (y/n/skip)`, then stop. On `y`, print the **Module N verbatim intro**.
  • **`next`**:
  • Read `.claude/memory/guide-progress.md`.
  • If missing, treat exactly like empty (first-run flow).
  • If progress exists, find the first uncompleted module and print its **verbatim intro** directly - NO "Ready?" gate. The user already said `next`; asking again is friction.
  • **A number 1-8** (e.g. `/guide 3`):
  • Read (or create) progress file.
  • Print one line: `Jumping to Module N - [title].`
  • Ask `Ready to start Module N? (y/n/skip)` and stop.
  • On `y`, print the **Module N verbatim intro**.
  • **`reset`**:
  • Overwrite `guide-progress.md` with a fresh copy (all modules unchecked, `started:` = today).
  • Print one line: `Progress reset.`
  • Then show the **First-run orientation block**.

The "Ready?" prompt at a gate is the ONLY prompt for that turn. Do not preview the intro on the same turn as a gate - that buries the gate.

Answers at a "Ready?" gate:

  • `y` - print the module's verbatim intro on the next turn
  • `n` - stop (one-line acknowledgement)
  • `skip` - mark the current module `[x]`, move to the next gate

---

First-run orientation block (print verbatim)

Welcome to **/guide** - an interactive tour of the shipped skills.

**What this is**: 8 short modules (~5-10 min each), one per shipped skill, run against sample data in `samples/`. Do 1-2 per session over days, not all at once. Progress is saved in `.claude/memory/guide-progress.md`, so you can stop and resume. The framing adapts to the role set in `CLAUDE.md`.

**The 8 modules**:
1. `/meeting` - raw meeting notes to structured note + tasks
2. `/today morning` - daily planning (mood, energy, focus items)
3. `/decision` - document a product decision with options + trade-offs
4. `/communicate` - draft a Slack / email / async update
5. `/create-epic` - raw idea to structured epic draft
6. `/competitive-research` - build a sourced competitive matrix
7. Capstone - chain `/meeting` then `/decision` then `/communicate`
8. Extend your workspace - run `/opportunity-solution-tree`, add or write more skills

**Commands**:
- `/guide` - continue / resume (or start at Module 1 the first time)
- `/guide next` - go to the next uncompleted module (no confirmation)
- `/guide <n>` - jump to a specific module (e.g. `/guide 3`)
- `/guide reset` - wipe progress and start over

**At a "Ready?" gate**: `y` start, `n` stop, `skip` mark done and move on.

Ready to start Module 1? (y/n/skip)

Print this block and STOP on the same turn. Do not append the Module 1 intro here - it's for the next turn, af

Read more
Read it on GitHub ↗

Showing the first part of this file.

Ships withai-pm-workspace

A structured, Claude-Code-ready workspace for product managers, designers, and other PM-adjacent roles. Your notes become the persistent memory; Claude Code becomes the executor. This is a template.

Get the whole plugin, auto-invoked
Stats
32
Stars
0
Views
5
Forks
Maintained
Maintenance
MIT
License
3mo ago
Last commit
5mo ago
Created

Repo: aleksander-dytko/ai-pm-workspace