Skip to content
Back to skills

Project Memory

ASecurity

Consult or maintain routed project memory without loading unrelated topics. Use for unfamiliar work or durable repository discoveries.

  • 14 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 19, 2026
developmentrustdebugginggit

Works with

  • cli

Security analysis

A100/100

Scanned October 19, 2026

npx -y skills add adrienm7/ergopti --skill project-memory --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Project Memory?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Project Memory
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/adrienm7-project-memory/badge)](https://www.skillsdirectory.com/skills/adrienm7-project-memory)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: project-memory
description: Consult or maintain routed project memory without loading unrelated topics. Use for unfamiliar work or durable repository discoveries.
---

# Project memory

`docs/memory/README.md` is the routing index for hard-won knowledge. The actual
entries live beside it in focused files under `docs/memory/`.
`docs/PROJECT_MEMORY.md` is only a compatibility pointer for older clients.
This catalog is shared by every developer, agent and reviewer and replaces
agent-private memory stores.

## Consult it first

Read `docs/memory/README.md`, then only the topic files covering the area you
will touch. Do not load every topic by default. A meaningful share of "new"
bugs in this repo are documented recurrences.

## What belongs here

Something **non-obvious**, **durable**, and written in **English**: a foot-gun that cost real debugging
time, an invariant that is not visible from the code, a decision whose rationale
would otherwise evaporate, an intentional asymmetry that looks like a bug.

What does **not** belong: anything the repo already records — code structure, the
content of a fix (that is the commit), git history, or universal rules already
written in `AGENTS.md`. Do not restate; link instead.

## Entry format

Add the entry to the narrowest existing topic file.

```markdown
### slug-with-dashes

One concise paragraph naming the mechanism and the action future work must take.
```

Use dash-separated slug headings. Keep entries short enough to scan and link
directly to code/tests when that is more useful than restating them.

## Writing well

- **Be concrete.** "Guard the map access" is useless; name the file, the key, the
  failure mode, and what the symptom looked like.
- **Convert relative dates to absolute.** "Last week" is meaningless in six months.
- **Record the symptom, not just the cause** — that is the string a future reader
  searches for when they hit it again.
- **Check for an existing entry first** and extend it rather than adding a near
  duplicate. Delete entries that turn out to be wrong.
- **Delete stale history.** Completed audit transcripts, branch narratives,
  commit summaries, temporary measurements, and TODO snapshots belong in Git,
  not project memory.
- **Verify before trusting.** Entries reflect what was true when written. If one
  names a file, function or flag, confirm it still exists before acting on it —
  and correct it if not.

## Cadence

Add or update entries proactively at the end of a task, without being asked.
Prune obsolete material at the same time. Put measured rejected ideas in
`docs/memory/rejected_proposals.md`, not among implementation invariants. Then
commit without pushing, per the `commit-and-push` skill.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.