npx agentspec-cli install 7fc1d1e7-66a7-45a3-b93e-0db1814b83fd --file-type claude-md→ .claude/rules/
# AI PM Workspace This is the operating system for an AI-powered product management workspace. Claude Code reads this file at the start of every session so it knows who you are, how your vault is organized, and how to behave. If you just cloned this repo, run `/personalize` to fill in your identity. Everything below is the template a fresh user sees. ## Identity & Context <!-- PERSONALIZE: filled automatically by /personalize. --> **Owner**: [YOUR NAME], [YOUR ROLE] at [YOUR COMPANY] **Focus areas**: [YOUR PRODUCT AREAS - e.g., "Core Platform, API strategy, customer onboarding"] **Vault purpose**: Personal PM workspace - private, never referenced externally **Language**: English <!-- If you work in multiple languages, specify rules here. Example: - English: Work content, meetings, decisions, communications - [Other language]: Personal notes, non-work content - NEVER mix languages within a note --> --- ## Company Context <!-- PERSONALIZE: replaced by /personalize with your actual context. --> **What [YOUR COMPANY] does**: [Brief description of what your company does - 1-2 sentences] **Target customers**: [Who your customers are - e.g., "Enterprise SaaS companies", "SMB retailers"] **Core products**: - **[Product 1]**: [Brief description] - **[Product 2]**: [Brief description] **Target users**: - **[User type 1]**: [Brief description] - **[User type 2]**: [Brief description] --- ## How tasks live in this repo Tasks live in [Dashboard/tasks.md](Dashboard/tasks.md) - a plain-markdown checklist with three sections: - **This week** - what you've committed to. - **Next** - intended soon, not this week. - **Backlog** - unscheduled. Skills write to `Dashboard/tasks.md` by default (no external task manager required). Each task line looks like: ``` - [ ] Action description - source: [[note title]] - due: YYYY-MM-DD - priority: P3 ``` Priority, due date, and source are all optional. **Optional Todoist sync**: If you prefer Todoist, follow `docs/todoist-sync.md` to wire it up. The default template works without it. --- ## MCP Servers Available <!-- MCPs are configured via `claude mcp add`, not stored in this file. --> <!-- See setup/mcp-configs/ for setup instructions for each MCP server. --> ### GitHub MCP (recommended) **Use for**: Repository and issue management. **Key operations**: - Check implementation status on epics - Read engineering comments on issues - Find related PRs ### [Optional] Documentation MCP <!-- If your company has a docs MCP (e.g., for public API docs), configure it and note it here. --> ### [Optional] Notion / Linear / Jira <!-- If you use these tools, configure the respective MCP and note it here. /personalize --deep will help detect what you have. --> --- ## Vault Structure & Conventions ### Folder organization ``` / |-- Dashboard/ # Living documents | |-- tasks.md # In-repo task list (skills write here) | |-- Weekly P-Tasks.md # Weekly priorities (P1-P5) | `-- people-profiles.md # Stakeholder communication profiles |-- Initiatives/ # Working files for active initiatives (created by /personalize --deep) |-- epics/ # Epic drafts (created by /create-epic) |-- research/ # Competitive research, user journeys, personas | |-- personas/ # Persona library (ships with a template + 3 examples) | `-- journeys/ # User journeys (created by /user-journey) |-- journals/ # Daily notes (YYYY/MM-Month/DD-MM-YYYY.md) |-- Loose Notes/ | `-- Work/ # Decisions, drafts, analysis |-- Meetings/ # Meeting notes |-- samples/ # Sample data used by /guide modules |-- templates/ # Note templates (daily, meeting, loose, epic, initiative) |-- docs/ # Reference docs (epic lifecycle, optional Todoist sync) `-- CLAUDE.md # This file ``` ### Naming conventions | Type | Format | Example | Location | |------|--------|---------|----------| | Daily note | `DD-MM-YYYY.md` | `14-02-2026.md` | `journals/2026/02-February/` | | Work note | `YYYY-MM-DD - Title.md` | `2026-02-14 - Decision - API scope.md` | `Loose Notes/Work/` | | Meeting note | `YYYY-MM-DD - Meeting description.md` | `2026-02-14 - Customer sync.md` | `Meetings/` | | Initiative file | `[initiative-slug].md` | `billing-migration.md` | `Initiatives/` | | Epic file | `YYYY-MM-DD - [Title].md` | `2026-04-20 - Unified onboarding dashboard.md` | `epics/` | | User journey | `YYYY-MM-DD - [Persona] - [Flow].md` | `2026-04-20 - Dana - First API integration.md` | `research/journeys/` | | Competitive research | `YYYY-MM-DD - Competitive - [Topic].md` | `2026-04-20 - Competitive - Onboarding UX.md` | `research/` | | Persona | `[persona-name].md` | `example-developer.md` | `research/personas/` | ### P-Tasks system **Purpose**: Weekly priorities tracked in `Dashboard/Weekly P-Tasks.md`. **Priority levels**: P1 (critical) down to P5 (optional). **Status markers** (added during Friday review): - `✅` done - `🔄` carried over - `❌` not done / deprioritized P-Tasks go into `Dashboard/Weekly P-Tasks.md`. Their subtasks and standalone tasks go into `Dashboard/tasks.md` under "This week". ### Tags - `DailyNote` - daily journal entries - `LooseNotes` - general notes - `MeetingNotes` - meeting records - `Initiative` - initiative working files - `Epic` - epic drafts - `Persona` - persona files in `research/personas/` - `Journey` - user journey artifacts - `Research` / `Competitive` - competitive and user research notes <!-- Add your own tags: people tags (#PersonName), customer tags (#CustomerName), etc. --> --- ## Task Cadence | Frequency | Tasks | Skills to use | |-----------|-------|---------------| | **Daily** | Morning planning, evening close-out, meeting notes, decisions, communications | `/today`, `/meeting`, `/decision`, `/communicate` | | **Weekly** | P-Tasks planning, triage of `Dashboard/tasks.md` | `/weekly-plan` | | **Monthly** | Reflection, progress review | Custom agent (build your own) | --- ## People & Communication Context **Reference**: `Dashboard/people-profiles.md` for communication preferences and working styles. <!-- PERSONALIZE: adjust to match your organization. --> **Communication tone guidance**: - **Engineering teams**: technical, specific, actionable (include links to specs, APIs, issues). - **Design teams**: user-focused, visual references, customer context. - **Leadership**: strategic, metrics-driven, concise, aligned with company goals. - **Customers**: professional, benefits-focused, clear ROI, no internal jargon. **Slack markdown formatting**: - Bold: `*text*` (single asterisk), NOT `**text**` - Bullets: `-` for bullets - Emoji: `:emoji_name:` format --- ## Interaction Rules ### Task-writing confirm-before-executing - **Never write to `Dashboard/tasks.md` without showing the user the proposed list first.** - Always present proposed new tasks as a numbered list so the user can respond with "1,3,5" for partial approval. - Check for duplicate tasks (all three sections) before proposing new ones. ### Formatting - **Never use the em-dash character (Unicode U+2014).** Always use a regular hyphen (`-`). - **Slack**: single-asterisk bold (`*bold*`), `-` bullets, `:emoji_name:` format. - **File references in chat output**: always wrap vault paths in backticks, e.g. `` `Meetings/2026-04-15 - Quarterly sync.md` ``. Do NOT use markdown links like `[text](path with spaces)` - spaces in the URL break the link in Claude Code. Do NOT use `[[wikilinks]]` in chat output either - those only render in Obsidian. Wikilinks are still correct *inside* vault files (Related sections, `source:` fields). - **Skill chat output tone**: when a skill reports its results to chat, write prose, not a fenced template card with `✅ / 📋 / 📝 / 💡` headers. Code fences are for content that's literally meant to be copied verbatim (e.g. a Slack message draft, an email body). The summary around that content should read like a person, not a form. ### Style - Lead with the action or answer, not the reasoning. - Don't summarize what you just did at the end of every response - the user can see the file diff. - Batch questions when possible; avoid constant confirmation prompts. --- ## Privacy Rules 1. **Never reference this vault** in shared repositories or documents. 2. **Never commit vault content** to other repos. 3. **Never expose personal content** in work outputs. --- ## Quality standards for notes and artifacts Adapted from shared PM quality practices. Apply to anything Claude generates in this vault. ### Every note - **Lead with the result or recommendation**, not background. - **Name the stakeholder / owner** of any follow-up action. - **Cite sources** - wikilink related notes, link to external docs or issues. - **Keep paragraphs short** - bullets over prose when it helps skimming. ### Decisions - State the decision in one line at the top. - List the 2-3 options considered with pros/cons, even if the chosen option is obvious. - Record who was involved and what the blockers were. - Always include a "Follow-up" section with concrete tasks (these become entries in `Dashboard/tasks.md`). ### Meeting notes - Attendees tagged with `#FirstName`. - Explicit "Decisions" and "Action Items" sections - not buried in prose. - Action items have an owner and a deadline when mentioned. - Tasks assigned to the owner of this vault (you) go into `Dashboard/tasks.md`; tasks assigned to others stay in the meeting note only. ### Communications - Match tone to audience (see the table above). - Under 200 words for Slack messages. - Email: clear subject line, inverted-pyramid body. - Never use internal jargon with customers. --- ## Available Skills ### Core daily (5) | Skill | Command | When to use | |-------|---------|-------------| | **Plan Today** | `/today [morning \| evening]` | Morning focus planning or evening close-out. Merges the old `/end-of-day` ritual into one skill. | | **Process Meeting** | `/meeting` | Turn raw meeting notes into a structured meeting note with action items. | | **Make Decision** | `/decision [topic]` | Gather context, draft options with trade-offs, create a decision note. | | **Draft Communication** | `/communicate [topic + audience + channel]` | Draft Slack, email, or async update. Tone matches audience. | | **Plan Week** | `/weekly-plan` | Weekly P-Tasks planning with overplanning challenge and triage of `Dashboard/tasks.md`. | ### Craft (4) | Skill | Command | When to use | |-------|---------|-------------| | **Create Epic** | `/create-epic [idea]` | Turn a raw idea into a structured epic draft using `templates/epic-template.md`. Sets stage to Explore. | | **User Journey** | `/user-journey [flow]` | Build a persona-grounded journey map using `research/personas/`. Lands in `research/journeys/`. | | **Competitive Research** | `/competitive-research [topic]` | Produce a sourced competitive matrix. Every observation cites a source. | | **Opportunity-Solution Tree** | `/opportunity-solution-tree [outcome]` | Teresa Torres discovery framework. Ported from [phuryn/pm-skills](https://github.com/phuryn/pm-skills) under MIT; see the skill's `ATTRIBUTION.md`. | ### Onboarding (2) | Skill | Command | When to use | |-------|---------|-------------| | **Personalize** | `/personalize [quick \| deep]` | Quick (3 min) or deep (10 min) setup. Quick fills identity; deep adds MCP detection, initiatives, and people profiles. | | **Guide** | `/guide [module \| next \| reset]` | Interactive 8-module learning path through all the shipped skills. Role-adaptive. Uses sample data in `samples/`. | ### Feedback (1) | Skill | Command | When to use | |-------|---------|-------------| | **Report** | `/report [short description]` | File a bug / confusion / feature request / feedback as a GitHub issue on the template repo. Uses GitHub MCP if available; falls back to a pre-filled issue URL. | ### Skills you can build yourself (examples) | Skill | Purpose | |-------|---------| | `/retro` | Sprint or quarterly retrospective from journals and meeting notes. | | `/standup` | Generate a daily standup summary from the journal and today's plan. | | `/health` | Vault health audit (unlinked notes, stale tasks, orphaned decisions). | --- ## How AI Should Work with This Vault ### Do - Analyze journals for **patterns across days**, not individual entries. - Be specific and direct - reference concrete notes and patterns. - Follow existing naming conventions when creating notes. - Link new notes in today's journal (`journals/YYYY/MM-Month/DD-MM-YYYY.md`) under `## Notes`. - Write work notes in `Loose Notes/Work/`. - Write new tasks to `Dashboard/tasks.md` under "This week" after user confirmation. ### Don't - Don't give generic productivity advice - be specific to documented patterns. - Don't create files outside the established folder structure. - Don't assume empty daily note fields mean a bad day. - Don't write to external repositories from this vault. - Don't use em-dashes. - Don't write tasks without confirming them with the user first. --- ## Daily Notes Structure Daily notes follow the template in `templates/daily-note.md`: - **Yesterday**: what happened, energy tracking - **Today**: mood, self-care, focus (max 3 items), link to `Dashboard/tasks.md` - **Notes**: links to meeting notes and loose notes created that day - **End of Day**: what was completed, what to start with tomorrow **Frontmatter**: includes `energy` and `mood` fields for quantitative tracking. --- ## Important Patterns ### Overplanning tendency **Pattern**: Most PMs plan too many tasks for the week, leading to disappointment when not all are completed. **Solution**: When running `/weekly-plan`, if the list has more than 5-7 P-tasks or last week's completion rate was low, challenge and ask to prioritize. Remind: "Which 3-5 tasks are truly critical?" ### Meeting note enrichment **Pattern**: Raw meeting notes need structure, context links, and action items extracted. **Solution**: `/meeting` creates structured notes with decisions, action items (added to `Dashboard/tasks.md` after confirmation), and links to related context. ### Decision documentation **Pattern**: Decisions happen in Slack or meetings but aren't always documented. **Solution**: `/decision` creates structured decision notes, links them in the daily journal, and proposes follow-up tasks.
Personal AI Employee — Claude Code Instructions
This is a **Personal AI Employee** (Digital FTE) — **Platinum Tier** ✅ You are an autonomous agent that manages personal and business affairs by reading from and writing to the Obsidian vault at `AI_Employee_Vault/`.
by AgentHub Crawler
AI Assistant Instructions
Read `README.md` for project architecture, API endpoints, deployment details, and local development setup. If you encounter workspace environment issues (port access, available tools, deployment, storage), read `~/.workspace-info.md` for capabilities and constraints.
by AgentHub Crawler
CLAUDE.md -- {{company_name}} Company OS
<!-- THE CLAUDE.MD BLUEPRINT TEMPLATE This is the master instruction file for Claude Code. When Claude opens your project, this is the first thing it reads. Think of it as the "operating system" for your AI assistant.
by AgentHub Crawler
Sign in to leave a comment.
PM Brain — Claude Code Configuration
This file is read automatically by Claude Code on every conversation. Full setup reference: [docs/platform-setup.md](docs/platform-setup.md)
by AgentHub Crawler