5,505,781 AI resources indexedWorking toward the world’s largest AI resource index.

PDF

Update CLAUDE.md

Update CLAUDE.md based on what was just discussed in conversation

Installable skillMITBy tta-labUpdated Oct 30, 2026
INTENDED FORClaude CodeInstall in a listed agent.

instructions.md

---
name: update-claude-md-0399847c2756
description: "Update CLAUDE.md based on what was just discussed in conversation"
category: command
claude-code: {}
---

# Update CLAUDE.md

## Overview

After a meaningful conversation — decisions made, patterns discovered, preferences stated — the user may ask you to capture what was discussed into the project's CLAUDE.md. This ensures future sessions benefit from today's context.

## Which CLAUDE.md?

Find the CLAUDE.md you own in the current repo:

- **In `~/clawd/`** — update your agent file (e.g. `yuki/CLAUDE.md`, `kestrel/CLAUDE.md`)
- **In any other repo** — update the project root `CLAUDE.md`

If unsure, ask which file to update.

## How to Update

1. **Review the conversation** — identify decisions, preferences, patterns, or conventions that were established
2. **Read the current CLAUDE.md** — understand what's already documented
3. **Add only what's new** — don't duplicate existing content
4. **Place it in the right section** — match the file's existing structure. Create a new section only if nothing fits.
5. **Be concise** — write it as an instruction, not a narrative. Future agents need to act on it, not read a story.

## What to Capture

- Workflow decisions ("always use X instead of Y")
- Tool preferences ("use og pr, not gh")
- Conventions discovered ("tests go in __tests__/ not test/")
- Architecture decisions ("we chose X because Y")
- Gotchas and warnings ("don't run Z, it breaks W")

## What NOT to Capture

- Session-specific details (current task, in-progress work)
- Anything already in CLAUDE.md (check first!)
- Speculative conclusions from one conversation
- Temporary workarounds (unless they'll persist)

## After Updating

- Show the user what you added/changed
- Commit if asked (follow the repo's commit conventions)

Discuss this resource

Platform guidance

Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.