Permanent Memory Auto-Archiving System
Paste the following prompt into your AI chat to install this skill:
Please follow https://skillhub.cn/install/skillhub.md to install @user_6bedba2d/ybdsjy.
About this skill
Problem
After an agent session ends, preferences, decisions, and project context are easily lost on restart. Accumulating everything in one large file creates bloat and makes retrieval hard. This skill addresses local, auditable long-term memory by separating distilled conclusions from raw conversational records, preserving append-only history without unbounded disk growth.
How It Works
- Diary layer: appends conversation content to
diary/YYYY-MM-DD.mdas raw archival material. - Memory layer: at session end, extracts explicit preferences, key project facts, technical decisions, and contacts into
MEMORY.md, constrained to four sections and never overwritten. - Archive layer: compresses diaries older than 30 days into
archive/YYYY-MM-DD.gz, reducing storage while keeping them searchable. - Recall flow: for questions like “what did we decide last time,” it searches
MEMORY.mdand recent diaries first, then falls back to compressed archives, citing sources. - Local constraints: operations are scoped to
~/.hermes/memory, using the Python standard library andgzip, with no external network service.
Boundaries
It is suitable for cross-session memory, persistent preferences, continuous project context, and decision traceability. Avoid storing plaintext passwords or secrets in the diary; use dedicated log systems or databases for high-volume logs and structured queries. Do not break the four-section structure of MEMORY.md, and do not manually edit .gz archive files.
Use Cases
- An engineer maintaining a local AI agent wants new sessions to load past preferences and record project decisions.
- A project lead running multi-turn discussions needs to persist technology choices in MEMORY.md to avoid re-confirming after restarts.
- An independent developer using a personal agent needs daily conversations archived locally and searchable for earlier decisions.
- A product engineer testing agent behavior needs to verify memory writes, 30-day gzip archiving, and search recall work as expected.
Best For
- Engineers building local AI agents who want to preserve user preferences and project context after sessions end.
- Developers maintaining multi-project workflows who need to trace technical decisions and contact information over time.
- Independent developers using personal assistant agents who want to persist conversation logs and archive them compactly.
- QA engineers testing agent reliability who need to verify memory writes, retrieval, and archiving behavior.
Related Skills
Collects daily conversations, task logs, and errors, then runs a nine-module review to surface blind spots, generate actions, and update memory.
Run a grilling session to interact with or test AI agents.
A systematic prompt optimization skill that refines prompts using a four-step distillation framework (diagnose, structure, think, compress) and methodologies from four prompting masters.
A meta-skill for the FlagOS skills repository to create, improve, validate, and evaluate agent skills, covering SKILL.md, scripts, references, and convention checks.