- Home
- Skills
- Agents & Orchestration
- project-memory
Project Memory 1
Maintain durable, repository-local memory and state between different AI coding agents and sessions.
Free
Works with the AI Agents you already use
project-memory
Example session with this skill installed
I'm starting work on the authentication module. Initialize the project memory and set a checkpoint for the current state.
- Read your context and instructions
- Compiled the project-memory
I have initialized the project memory files and recorded the starting state. The files have been added to your local git excludes to ensure they are not committed.
Project Memory Initialized
| File | Status | Purpose |
|---|---|---|
AGENTS.md | Created | Stable operational instructions and repo-specific rules. |
AI_CONTEXT.md | Created | Entry point, project purpose, and architectural constraints. |
docs/project-memory/CURRENT_STATE.md | Created | Current branch status and verification baseline. |
.git/info/exclude | Updated | Local-only memory files excluded from version control. |
Current Snapshot
- Start Time: 2025-05-20T14:30:05-0700
- Branch:
feature/auth-module - Head:
a1b2c3d - Status: 2 files modified (untracked)
Next steps
- Review
AI_CONTEXT.mdto ensure the authentication entry points are correctly identified. - Begin implementation of the login controller.
- A durable checkpoint will be created automatically in 15 minutes or at the next major task boundary.
Connects securely to your tools. The creator never sees your data.
About this skill
When switching between AI coding assistants like Claude Code and Cursor, or resuming work after a session timeout, context is lost. This skill solves the "AI amnesia" problem by maintaining a durable, repository-local memory that survives session resets and transitions between different LLM agents. It prevents agents from repeating mistakes, losing track of architecture decisions, or ignoring project-specific invariants.
What it does
- State persistence maintains a dated snapshot of Git branches, working tree status, and active tasks across sessions.
- Agent synchronization enables different coding tools to share the same context files, ensuring consistency in instructions and progress.
- Automated handoffs generates detailed recovery points including completed work, validation results, and exact resume commands.
- Local-only storage manages exclusions in
.git/info/excludeto keep AI-specific memory out of your repository's version history and Docker builds. - Project inventorying crawls local directories to map out existing repositories and their documentation status.
How it works
- Bootstrap the memory structure by reading existing READMEs and manifests to determine the repo's architecture and conventions.
- Initialize local-only files like
AGENTS.mdandAI_CONTEXT.mdto define operational rules and project entry points. - Record durable checkpoints every 15 minutes of work to ensure no progress is lost during long-running tasks.
- Handoff the session by documenting verified facts, open questions, and specific next steps for the next agent session.
Frameworks & tools
This skill is designed for Git-based workflows and works with any AI coding agent including Claude Code and Codex. It uses standard shell utilities and a local Python inventory script.
Why this beats prompting it yourself
Manual prompting lacks the persistence needed for multi-day tasks and cannot automatically sync state between different AI tools. This skill enforces a standardized schema for project memory, ensuring that every agent follows the same invariants and respects your local environment limits without manual re-explanation.
Use cases
- Resuming a complex refactor after an LLM context window reset.
- Handing off a partially completed feature from one AI agent to another.
- Onboarding an AI assistant to a large, undocumented legacy repository.
- Maintaining a local roadmap and decision log that isn't committed to the main branch.
Known limitations
Memory files are deleted by git clean -x commands. Local memory is checkout-specific and does not automatically follow the user to new git worktree instances without manual path naming.
Memory is lost on git clean -x commands. It exists only in the local checkout and is not automatically shared across git worktrees or remote clones.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 7 days ago
- Free with an Agensi account
Requires