- Home
- Skills
- Agents & Orchestration
- persona and memory
persona and memory
A file-based long-term memory and persona system with weighted decay and identity consolidation.
$9.99
Works with the AI Agents you already use
persona and memory
Example session with this skill installed
I've mentioned twice now that I prefer using tabs over spaces for all my Python projects. Please remember this for the future.
- Read your context and instructions
- Compiled the persona and memory
I have recorded your preference for tabs in Python projects. Since this is a repeat mention, I have updated the memory weight.
Memory Update: m-20240520-01
---
id: m-20240520-01
kind: preference
weight: 0.63
hits: 2
verified: 1
status: active
decay: 0.92
created: 2024-05-20
last_seen: 2024-05-20
---
User prefers using tabs instead of spaces for Python development. This has been confirmed across multiple turns.
Next steps
- Run
python scripts/memory_cycle.py cycleat the end of this session to update theINDEX.md. - Mention this preference once more to trigger the "Identity" promotion threshold (verified >= 2).
- Check
memory/identity.mdafter the next cycle to see the permanent rule.
Connects securely to your tools. The creator never sees your data.
About this skill
AI assistants often lose their personality or forget user preferences between sessions. They either hallucinate past interactions or reset to a generic state every time you start a fresh thread.
This skill implements a local, file-based memory system that mimics human cognition. It uses a weighted decay algorithm to ensure that important facts stick while irrelevant noise fades away. It separates "Voice" (how the agent speaks) from "Mind" (what the agent knows), preventing your system prompt from becoming a bloated mess of conflicting rules.
What it does
- Persistent Persona maintains a consistent voice, register, and flourishes via a dedicated
persona.jsoncard. - Weighted Memory assigns weights to every interaction, increasing importance with usage and decaying it over time.
- Automated Fading moves low-weight entries to a "fuzzy" state, replacing full text with one-line summaries to save context window.
- Identity Consolidation promotes facts that are verified twice and used frequently into a permanent
identity.mdfile. - Audit Trails keeps all archived and superseded memories as plain text files for manual review or retrieval.
How it works
- Initialize the memory structure using the provided Python script to create the persona card and entry directories.
- Recall by reading the
INDEX.mdat the start of a session to see what is currently "active" or "fuzzy". - Record new episodes or preferences by creating markdown files with specific frontmatter metadata.
- Maintenance runs at the end of a session to recalculate weights, apply decay, and fold old memories.
Frameworks & tools
The core logic uses a Python 3.9+ script with standard library components. It requires no external databases or network access, operating entirely on local Markdown and JSON files.
Why this beats prompting it yourself
Standard prompting for "memory" usually leads to context bloat or the agent "remembering" things that never happened. This skill uses deterministic arithmetic to handle forgetting and verification, ensuring the agent only claims to know what is actually written in the index.
Use cases
- Long-term Companions that develop a unique personality based on months of interaction.
- Coding Assistants that remember specific project preferences and "gotchas" without re-explaining them.
- Character Bots for roleplay that need to track complex relationship histories and world facts.
Known limitations
Does not support concurrent writes from multiple agents. Requires a host environment capable of reading/writing local files and executing Python scripts to automate the weight cycles. This skill is self-contained. If you also want the same persona living in Telegram or Discord — running unattended, with its own memory per chat — that part ships separately: https://github.com/1-dcjiang/fish-runtime/releases/latest
No concurrent write support. Requires Python 3.9+ for automated maintenance. Host must have filesystem access. Not suitable for per-repo technical notes.
This skill is self-contained. If you also want the same persona living in Telegram or Discord —
running unattended, with its own memory per chat — that part ships separately:
https://github.com/1-dcjiang/fish-runtime/releases/latest
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 6 days ago
Requires