- Home
- Skills
- Agents & Orchestration
- ctxmem agent skill
More screenshots
Works with the AI Agents you already use
ctxmem agent skill
Git-native local project memory for tracking architectural decisions and context across agent sessions.
Free
ctxmem agent skill
Example session with this skill installed
I'm starting the migration to the new auth provider. Can you check if we made any specific decisions about the JWT implementation in previous sessions?
- Read your context and instructions
- Compiled the ctxmem agent
I have queried the local project memory for decisions related to "JWT implementation" and "auth provider migration."
Project Memory Search Results
| ID | Type | Title | Summary | Path |
|---|---|---|---|---|
rec_01j8 | decision | JWT Algorithm Selection | Use RS256 for all service-to-service auth; avoid HS256. | docs/arch/auth.md |
rec_05x2 | decision | Token Expiry Policy | Set access tokens to 15m and refresh tokens to 7d. | src/auth/config.py |
rec_09y1 | correction | Auth Header Format | Use 'Bearer' prefix strictly; previous 'Token' prefix is deprecated. | src/api/middleware.ts |
The memory store indicates a strong preference for RS256 and specific expiry windows. I also found a correction record (rec_09y1) that supersedes the older header format used in the prototype. I will proceed with the migration using these constraints.
Next steps
- Review
src/auth/config.pyto ensure the 15m/7d expiry logic matches these records. - Initialize the migration script using the RS256 algorithm as recorded.
- Run
ctxmem synconce the initial boilerplate is committed to update the index.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop losing track of architectural decisions and project context between coding sessions. Developers often struggle with agents that forget previous discussions or hallucinate outdated patterns in growing repositories. This skill implements a Git-native memory layer that keeps your project knowledge where it belongs: in the repository.
What it does
- Persistent recall uses local keyword search to retrieve relevant project history before starting new tasks.
- Context management generates task-specific context within token budgets to prevent LLM noise.
- Decision tracking records architectural choices and rationales with a built-in superseding mechanism for updates.
- Git-native storage saves memories in JSONL format within your repo for easy versioning and sharing.
- State reconciliation identifies when current code conflicts with historical records to ensure accuracy.
How it works
- Inspection checks for the presence of the ctxmem CLI and repository configuration files.
- Setup installs the local utility and configures persistent agent instructions for automatic recall.
- Retrieval runs targeted queries against the local memory store to gather evidence for the current prompt.
- Recording captures new decisions or corrections using specific types and file path associations.
Frameworks & tools
This skill requires Python 3.8+ with SQLite FTS5 support and Git. It operates locally using the ctxmem CLI tool.
Why this beats prompting it yourself
Manual prompts rely on the agent's short-term context window which clears every session. This skill provides a structured, searchable database that persists across tool restarts and different developer environments.
Use cases
- Onboard new agents or developers to a complex codebase with historical context.
- Prevent regressions by surfacing previous architectural "why" decisions.
- Synchronize project knowledge across a team using standard Git workflows.
- Manage large-scale refactors where decisions must be tracked over multiple days.
Known limitations
Requires local execution permissions for Python and Git. Does not store secrets or credentials. Memory recall performance depends on the quality of recorded titles and descriptions.
Requires Python 3.8+, SQLite FTS5, Git, and local command execution with project read/write permissions. Automatic recall across sessions depends on the host loading and following persistent project instructions. Reload the agent session after setup. Network access is needed only for installation.
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
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
Requires