From Harness Base
Write, index and prune durable knowledge as markdown notes in .claude/memory/. Use when you learn a fact, make a decision, resolve an incident, or discover a recurring procedure that a future session would need.
How this skill is triggered — by the user, by Claude, or both
Slash command
/harness-base:memory-md-filesThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Durable knowledge lives as small markdown notes in `.claude/memory/`. This skill
Durable knowledge lives as small markdown notes in .claude/memory/. This skill
defines how to write, index and prune them. The goal: a future session can
recover what this session learned by reading files, not transcripts.
Write a note when something would be expensive to rediscover:
Do not write a note for transient task state — that belongs in progress
tracking. Do not duplicate CLAUDE.md; CLAUDE.md is for always-true facts.
Filename: NNNN-kebab-slug.md, where NNNN is the next free 4-digit id.
---
id: "0007"
type: semantic # semantic | episodic | procedural
title: "Short human title"
created: 2026-05-15
updated: 2026-05-15
tags: [auth, ci]
status: active # active | archived
---
# Short human title
**Context.** Why this came up.
**Decision / Finding / Procedure.** The substance.
**Consequences.** What this implies for future work.
**Links.** Related notes, files, PRs.
Keep each note under ~40 lines. One idea per note. Notes are cheap; make many.
.claude/memory/MEMORY.md is the index — the agent reads it at session start.
Every note must appear in it. After writing or editing a note:
MEMORY.md.Updated date.A note not in the index is invisible. Never skip this step.
If .claude/memory/MEMORY.md does not exist yet, create it from this skeleton:
# Memory index
The agent reads this file at session start, then opens the notes relevant to
the task. Keep one row per note. Newest at the top within each section.
## Semantic — what is known (decisions, facts, constraints)
| Note | Title | Updated |
|---|---|---|
## Episodic — what happened (incidents, debugging sessions, outcomes)
| Note | Title | Updated |
|---|---|---|
## Procedural — how things should be done (recurring runbooks)
| Note | Title | Updated |
|---|---|---|
## Archive
Stale or superseded notes are moved to `.claude/memory/archive/` and listed
here so history is not lost.
| Note | Title | Archived | Reason |
|---|---|---|---|
Memory rots. On any session that touches .claude/memory/:
status: archived,
move the file to .claude/memory/archive/, and move its index row to the
Archive table with a reason. Do not delete — provenance matters.updated.Aim to keep the active index short enough to scan in seconds.
npx claudepluginhub camilool8/harness-engineering-templates --plugin harness-baseGuides collaborative design exploration before implementation: explores context, asks clarifying questions, proposes approaches, and writes a design doc for user approval.
Creates structured, bite-sized implementation plans from specs or requirements before writing code. Useful for breaking down multi-step tasks into testable steps with file structure and task boundaries.
Resolves in-progress git merge or rebase conflicts by analyzing history, understanding intent, and preserving both changes where possible. Runs automated checks after resolution.