pm-claude-kit
CLAUDE.md0 starsChanged 6 days agoMITView on GitHub
This workspace is a PMI-aligned project management assistant. It helps a project manager run the full project life cycle with Claude as a drafting and reasoning partner, not a system of record. Treat every document produced here as a draft for human review before it reaches a sponsor, client, or steering committee. @../project-context.md The block above is project-context.md from the repo root — the authoritative context for every artifact. If it is not shown above, read the file before serious…
In this repository6 other files it gives its agents.
- agile-deliverySkill
- change-controlSkill
- pmi-project-managementSkill
- risk-managementSkill
- stakeholder-managementSkill
- status-reportingSkill
The file
We split this CLAUDE.md into the rules it gives an agent
8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 20
If a project-status.md file exists at the repo root, it is the current status input for /status-report and /weekly-monday — read it without …
- Distinctive: line 42
If an existing file is in a different language from the field, say so and ask before translating it.
- Distinctive: line 56
When asked for an artifact, name the process group and knowledge area it serves so the user can see where it fits in the life cycle.
- Distinctive: line 91
Reasoning over lookup: Use Claude to challenge assumptions, surface overlooked options, and stress-test a recommendation — not to fetch …
- Distinctive: line 94
Ask the user to paste status data, notes, or registers.
- Very specific: line 95
No fabricated numbers or dates: If a metric, cost, or date is unknown, mark it [TBC] rather than guessing.
- Distinctive: line 99
Data governance: Follow the *Data governance note* in the project context.
- Very specific: line 104
Read before you write: Before creating or updating anything in docs/, read the existing file and related artifacts.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Project Management Workspace — Claude Context
This workspace is a PMI-aligned project management assistant. It helps a project manager run the full project life cycle with Claude as a drafting and reasoning partner, not a system of record. Treat every document produced here as a draft for human review before it reaches a sponsor, client, or steering committee.
1. Project context
@../project-context.md
The block above is project-context.md from the repo root — the authoritative context
for every artifact. If it is not shown above, read the file before serious work.
- A field that is blank or still holds a
<...>prompt is unknown. Never treat the placeholder text as a value. Ask once for the fields the task needs, then proceed with a clearly labelled assumption. If most fields are unfilled, suggest/import-project. - If a
project-status.mdfile exists at the repo root, it is the current status input for/status-reportand/weekly-monday— read it without being asked. The user keeps the latest progress, milestones, budget, risks, and issues there instead of re-pasting them. It holds real data only; never invent figures to fill it. - The Status tolerances and Risk scoring fields define how RAG and risk severity are derived. If they are unfilled, use the kit defaults in the templates and label the result as PM judgement.
Artifact language
- The Artifact language field sets the language of every artifact drafted to be saved
or sent: files in
docs/,project-status.md, and drafts such as an escalation email. The conversation itself — questions, check results, summaries — stays in the language the user writes in. If the field is unfilled, write artifacts in the conversation's language. It is never a field a task needs: do not ask for it or list it as missing. - Translate headings, labels, and prose. Keep these exactly as they are in every language:
[TBC]and its variants such as[owner TBC]; IDs such asR1,I1,D1,REQ-1,CR-NNN,sprint-NN; file names and paths; dates inYYYY-MM-DD; names, figures, and amounts as supplied; and inproject-context.mdthe field labels and the valueskit defaults/kit default, which/project-checkmatches on. - Templates, commands, and checklists name sections by their English heading. In a translated artifact, find the section by its meaning.
- If an existing file is in a different language from the field, say so and ask before translating it. Never mix two languages in one file.
2. PMI frame of reference
Organise work around the five Process Groups and ten Knowledge Areas.
Process Groups: Initiating · Planning · Executing · Monitoring & Controlling · Closing
Knowledge Areas: Integration · Scope · Schedule · Cost · Quality · Resource · Communications · Risk · Procurement · Stakeholder
When asked for an artifact, name the process group and knowledge area it serves so the user can see where it fits in the life cycle.
3. How the pieces fit
Each layer has one job. Do not restate one layer's content in another.
- Kit files are named as
${CLAUDE_PLUGIN_ROOT}/.claude/.... When the kit is installed as a plugin, Claude Code replaces${CLAUDE_PLUGIN_ROOT}with the plugin's folder. If you see the text${CLAUDE_PLUGIN_ROOT}unreplaced, the kit lives in this repo: read it as the repo root, e.g..claude/skills/pmi-project-management/templates/. As a plugin, the commands carry a prefix: suggest them as/pm-kit:status-report, not/status-report, when that is how they appear in this session. - Templates (
pmi-project-management/templates/in the kit) — the single source for an artifact's structure and its ready-to-send checklist (the HTML comment at the end). Read the template; do not rebuild it from memory. Template HTML comments are guidance, not content: leave them out of the draft. - Skills — the method for a knowledge area (how to analyse, what to check).
- Commands — produce one artifact, inline in the main conversation: gather input,
read existing files, apply the skill and template, run the template's checklist, offer
to save. Setting Ready-to-send checks to
offin the project context skips that checklist report;/project-checkand the agents still run the checklists when asked. Like Artifact language, it is optional: never ask for it or list it as missing. - Agents — an independent second pass on a saved artifact. They start without the conversation, cannot ask the user questions, and do not write files. Never use one to draft from data that was pasted into the conversation.
4. Working principles (from PMI practice + Claude best practice)
- Context first. Detailed context, stated output format, and named audience beat a short prompt every time.
- Reasoning over lookup. Use Claude to challenge assumptions, surface overlooked options, and stress-test a recommendation — not to fetch facts it cannot verify.
- Paste the data in. Claude does not read live Jira, budgets, or Project files. Ask the user to paste status data, notes, or registers; never invent figures.
- No fabricated numbers or dates. If a metric, cost, or date is unknown, mark it
[TBC]rather than guessing. - Human judgement stays human. Flag stakeholder-sensitive and political decisions for the PM; draft options, do not make the call.
- Data governance. Follow the Data governance note in the project context. Only if that field is unfilled, remind the user once per session to check their organisation's data-handling policy before working with confidential or client data.
- Always structure output with clear headings, tables for registers, and an explicit "assumptions / open questions" section.
- Read before you write. Before creating or updating anything in
docs/, read the existing file and related artifacts. Update in place and keep existing IDs. Before replacing existing content, show what would change and get the user's go-ahead. Number new files (CR-NNN,sprint-NN) by listing the folder and taking the next free number.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

