agentleFS
kaushikhegde11kaushikhegde11/obsidian-project-memory/project-memory/SKILL.md

project-memory

Skill0 starsChanged 2 months agoMITView on GitHub

Create and maintain per-project "memory" notes in the user's Obsidian vault. Each project gets its own folder with four files — README.md (stable overview), STATUS.md (current snapshot), progress.md (dated diary), Decision.md (decisions + rationale). The storage folder is configurable — no fixed vault structure. Use whenever the user says "project memory", "store memory about my project", "set up project notes", "track this project in Obsidian", "make a README/STATUS for <project>", or pastes project context to file. Companion skill: update-project-memory updates these files later. One project per folder — never mix two projects in one file.

24 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. 0 are common, 6 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 19 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If no config file is found or it is incomplete, run a short setup:

    line 30
  2. Very specific:

    Ask for the vault root. You may reuse ~/.claude/skills/obsidian-notes-skill/vault-config.yaml vault_root if it exists.

    line 31
  3. Distinctive:

    Ask which folder should hold project memory.

    line 32
  4. Very specific:

    Ask the date format (default DD-MM-YYYY).

    line 33
  5. Very specific:

    Create <projects_folder> if it does not exist.

    line 40
  6. Distinctive:

    Create all four on first setup.

    line 45
  7. Distinctive:

    Context: Ask the user to describe the project, or read the context they pasted.

    line 108
  8. Distinctive:

    Missing context: If a required README.md field (what / why / who / how AI should help) or the current status is missing and you cannot …

    line 109
  9. Very specific:

    Create an extra file or subfolder only when the project clearly needs it — for example architecture.md, links.md, or a research/ subfolder.

    line 122
  10. Distinctive:

    When context is missing, ask.

    line 128

“Common” is measured: kept by at least 6 in 10 of the 19 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Project Memory

Store durable, structured memory about a single project in the Obsidian vault. An AI assistant (or the user) reads these files later to recover full context fast.

This skill creates a project's memory. Its companion, update-project-memory, updates it.

Tool-neutral: it works in any coding agent that can read and write files — Claude Code (as a skill), OpenAI Codex (as a /project-memory prompt), or any other assistant with vault access. Below, "you" means whichever assistant is running these instructions.

Prerequisite: vault access

This skill assumes you already have read, create, and edit access to the user's Obsidian vault — through the Obsidian MCP server, plain filesystem tools, or your own sandbox mount. This skill does not set up that access. If you cannot read or write the vault, stop and ask the user to grant access first. In a sandboxed agent (Codex's default is workspace-write on the current directory only), the vault path usually sits outside the sandbox: ask the user to add it as a writable root, or fall back to printing each file as a fenced code block for the user to paste.

Configuration (no fixed folders)

This skill hardcodes no vault structure. It reads project-memory-config.yaml. Resolve the config path in this order and use the first that exists:

  1. $PROJECT_MEMORY_CONFIG — explicit override.
  2. ~/.config/project-memory/config.yaml — canonical, tool-neutral location.
  3. ~/.claude/skills/project-memory/project-memory-config.yaml — Claude Code install.
  4. ~/.codex/project-memory-config.yaml — Codex install.

Fields:

  • vault_root — absolute path to the vault.
  • projects_folder — the folder that holds each project's memory folder. Any folder the user chooses (for example Projects memory, 2. Source Material/Projects, Areas/Projects, or "" for the vault root).
  • date_format — date style for the logs.

If no config file is found or it is incomplete, run a short setup:

  1. Ask for the vault root. You may reuse ~/.claude/skills/obsidian-notes-skill/vault-config.yaml vault_root if it exists.
  2. Ask which folder should hold project memory. The user may pick or type any folder.
  3. Ask the date format (default DD-MM-YYYY).
  4. Save the answers to ~/.config/project-memory/config.yaml, creating the folder if needed.

project-memory-config.example.yaml in the repo documents every field.

Location and rule

  • Base path: <vault_root>/<projects_folder>/<Project Name>/.
  • Create <projects_folder> if it does not exist.
  • One project = one folder. Never put two projects in the same file or folder.
  • The folder name is the project name (keep it readable; spaces are fine).

The four core files

Create all four on first setup.

1. README.md — stable overview (changes rarely)

# <Project Name>

## What it is
<one or two short paragraphs>

## Why it matters
<the problem it solves / the value>

## Who it is for
<target users or audience>

## Useful links
- <url> — <label>

## How AI should help
- <clear, durable guidance for an assistant working on this project>

2. STATUS.md — current snapshot

# <Project Name> — Status

_Last updated: <DATE>_

## Where it stands
<current state in a few lines>

## Next action
- [ ] <the next concrete step>

## Blockers
- <blocker, or "none">

## Needs review
- <anything awaiting a decision or review, or "none">

3. progress.md — dated diary (newest entry on top)

# <Project Name> — Progress log

## <DATE>
- What happened: <...>
- What changed: <...>
- Next: <...>

4. Decision.md — decision record

# <Project Name> — Decisions

## <short decision title> — <DATE>
- Decision: <what was decided>
- Why: <the reasoning>
- Revisit: <when or what should trigger a re-look>

Intake (ask before writing)

  1. Project name. It becomes the folder name.
  2. Context. Ask the user to describe the project, or read the context they pasted. Pull out: what it is, why it matters, who it is for, links, current status, recent progress, and any decisions with reasons.
  3. Missing context. If a required README.md field (what / why / who / how AI should help) or the current status is missing and you cannot infer it safely, ask the user. Do not invent it.

Create

  • Folder does not exist → create it and write all four files from the context.
  • Folder already exists → this is an update. Use the update-project-memory skill instead.

Sorting content into the right file

  • Stable facts (what/why/who/links/AI guidance) → README.md.
  • Present state, next step, blockers, review items → STATUS.md.
  • Anything that happened on a date → progress.md.
  • A choice plus its reason → Decision.md.

Extra files (keep it minimal)

Create an extra file or subfolder only when the project clearly needs it — for example architecture.md, links.md, or a research/ subfolder. Do not add structure the project does not need. The four core files are the default.

Rules

  • Assumes vault read/create/edit access (see Prerequisite).
  • One project per folder. Never mix projects.
  • Date every progress.md and Decision.md entry in the configured format.
  • When context is missing, ask. Do not fabricate status, decisions, or history.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.