# Project Memory

> Query this project's cold memory when the user refers to previous work, prior decisions, repeated failures, historical rationale, or a known fragile path. Do not use for ordinary fixes answerable from current files and tests.

- Skill: `candycong2001-ui/project-memory` (Agent Skill)
- Install (CLI): `npx skillmds@latest add candycong2001-ui/project-memory`
- Raw SKILL.md: https://api.skillmd.com/api/skills/candycong2001-ui/project-memory/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: candycong2001-ui (https://skillmd.com/u/candycong2001-ui)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/candycong2001-ui/project-memory

---


# Project Memory

## Read

1. Confirm the current Git root matches this project.
2. Check current files and tests first.
3. If history is needed, run `pjm --help` before using an unfamiliar command.
4. Use the smallest useful query, usually `pjm brief` or `pjm search <query>`.
5. Verify memory-derived claims against current evidence before acting.

## Write

- Do not write long-term memory unless the user explicitly asks to preserve it.
- Put uncertain observations in a candidate or note with their source.
- Record decisions, useful failed attempts, and stable project facts, not activity streams.
- Never save secrets or private raw logs.

Projectmem is cold memory. Current repository evidence remains the truth source.