mnemos

Persist structured working memory across context compaction and session changes.

Updated Jun 19, 2026
One-click install
npx skills add https://github.com/lciacci/tessera --skill mnemos-lciacci
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mnemos
Source: https://github.com/lciacci/tessera/tree/main/skills/mnemos
Command: npx skills add https://github.com/lciacci/tessera --skill mnemos-lciacci

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Mnemos prevents context compaction, session changes, and agent fatigue from causing important goals, constraints, decisions, results, and handoffs to be forgotten or re-derived.

Core Features & Use Cases

  • Typed Working Memory: Stores goals, constraints, results, context, and checkpoints with distinct persistence and eviction policies.
  • Fatigue Monitoring: Measures token utilization, scope scatter, re-read behavior, and tool-call errors to trigger consolidation or checkpointing.
  • Session Recovery and Analysis: Restores task state after compaction or interruption, ingests Claude transcripts, and reports haziness, corrections, and action divergence.
  • Use Case: During a long coding task, use Mnemos to preserve architectural constraints and completed work, checkpoint before context compaction, and resume with the relevant state already available.

Quick Start

Initialize Mnemos, create a goal and constraints, and checkpoint the task state before continuing development.

Frequently Asked Questions about mnemos

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent context compaction from losing critical constraints during a long coding task?

To prevent context compaction from losing critical constraints, use typed working memory to store goals and checkpoints with distinct persistence policies, ensuring important state survives session changes.

How does agent fatigue monitoring work in long-running coding sessions?

Agent fatigue monitoring measures token utilization, scope scatter, re-read behavior, and tool-call errors to trigger consolidation or checkpointing before context is lost during long-running coding sessions.

What is the best way to recover structured working memory after a session interruption?

The best way to recover structured working memory after a session interruption is to restore task state from persisted checkpoints, which preserves architectural constraints and completed work.

Can I ingest Claude Code transcripts for agent-behavior analysis and haziness scoring?

Yes, you can ingest Claude Code transcripts to analyze agent behavior, reporting haziness, corrections, and action divergence by applying structured telemetry to the session history.

Does session recovery require typed graph storage and Claude Code hooks?

Yes, session recovery requires typed graph storage, Claude Code hooks, checkpoint persistence, and structured telemetry to execute selective eviction, restoration, and haziness scoring.

When should I trigger checkpoint persistence to avoid re-deriving decisions?

You should trigger checkpoint persistence before context compaction or when fatigue monitoring detects high token utilization, preventing important goals and decisions from being forgotten or re-derived.