Skip to content

RUNNING-DOCUMENT.md

Latest commit

 

History

History
175 lines (109 loc) · 4.24 KB

RUNNING-DOCUMENT.md

File metadata and controls

175 lines (109 loc) · 4.24 KB

Running Document

Read this at the start of every Cowork session — before anything else.

This is your continuity artifact. Without it, Claude starts every session with no memory of who you are, what you've decided, or what rules apply. With it, Claude picks up where you left off.

Session start prompt:

"Read RUNNING-DOCUMENT.md before we begin. Confirm you've loaded it and tell me what's most important to hold from it."

Partnership Status

Last updated: [DATE] Sessions completed: [NUMBER] Active since: [DATE]

Who I Am — Human Profile

Tell Claude who you are. Be specific. This is what Claude needs to not have to guess.

Name: [Your name] Age: [Age] Background: [Professional background — what you do, what you know] Technical level: [Non-technical / some technical / technical] Time constraints: [e.g., "Limited laptop time. Mobile-accessible most of day."] How I work best: [e.g., "Direct feedback. Short responses. No padding."] What I'm trying to achieve: [Your goals in plain language]

Active Projects

Project Status Priority Next Action
[Project name] Active / Paused High / Medium / Low [What needs doing]

Rules Claude Must Follow

These apply in every session without exception.

  1. [e.g., "Direct answer first. Context second."]
  2. [e.g., "Tell me when something is wrong before you fix it."]
  3. [e.g., "Never soften feedback to spare my feelings."]
  4. [e.g., "Ask one question maximum per response."]

Mode Commands

Say these to instantly change how Claude operates.

Command What it does
"Just answer." One sentence only
"Simplify." Plain language, no jargon
"Proceed." Start the work now
"Stop." Halt everything
"Truth check." Tell me what you know vs. what you're inferring
[Add your own]

What Claude Knows About This Project

Key context Claude needs. Update this as the project develops.

[Paste key facts, decisions, and context that Claude must hold across sessions]

Decisions Log

Record decisions here so they don't need to be re-explained or re-made.

Date Decision Rationale Status
[DATE] [What was decided] [Why] Active / Superseded

Corrections Log

When Claude gets something wrong, log it. Patterns here prevent repeated errors.

Date What went wrong Correction Prevention rule
[DATE] [Error] [Fix] [Rule to prevent recurrence]

Open Questions

Things that are unresolved and need answering. Claude should raise these if relevant.

  • [Question — why it matters]
  • [Question — why it matters]

Active Ideas

Ideas in progress — not decisions yet, but worth holding.

  • [Idea and brief context]
  • [Idea and brief context]

Files In Play

Which files are authoritative. Claude references these — not older versions.

File Status Notes
[filename] Active / Frozen / Deprecated [Brief note]

Scheduled Tasks Running

Automated tasks — Claude should be aware these are running.

Task When Purpose
[Task name] [Schedule] [What it does]

Next Session Starts Here

Leave a clear note. Future Claude will read this cold.

Pick up from: [Exactly where to resume] Most important thing to hold: [The one thing that must not be forgotten] Open decision needing resolution: [Anything waiting for a call]

Session Log

Brief notes from recent sessions. Prune when older than 4 weeks.

[DATE] — Session [N]

  • [What happened]
  • [What was decided]
  • [What's next]

What I'm Watching

Things in the world relevant to this work. Claude should connect these when relevant.

  • [Topic — why it matters to the project]
  • [Topic — why it matters to the project]

Remember

The Running Document is not bureaucracy. It is the memory of this collaboration.

If it's stale, Claude is guessing. If it's current, Claude is a partner.

Update it. It takes two minutes. It saves hours.