Compress long conversations into structured memory indexes under .codex-memory.
npx skills add https://github.com/10kseason/Deep-memory-skills --skill deep-memory
Please help me install this Agent Skill. Skill: deep-memory Source: https://github.com/10kseason/Deep-memory-skills/tree/main Command: npx skills add https://github.com/10kseason/Deep-memory-skills --skill deep-memory
Turns long conversations and project logs into a compact, durable memory index that preserves decisions, commands, milestones, and critical context while filtering out noise.
Create a raw handoff note under .codex-memory/<project-slug>/raw, then run compressor.py to generate a durable compressed snapshot.
High-intent search queries and answers about installing and using this skill.
To preserve project state across multiple coding sessions, you can compress long conversation logs into a durable memory index that stores decisions, commands, and milestones under a project-specific directory for targeted recall without reloading full transcripts.
The best way to compress long conversation logs into structured memory is to create raw handoff notes and run a built-in compressor script, which converts the raw notes into structured snapshots and renders compact summaries for cross-session continuity.
You need durable memory management for long-running software projects when you must recall past architecture decisions, bug histories, file paths, metrics, and open loops across multiple sessions without reloading the full original transcripts.
No, this conversation memory tool requires no external dependencies to manage session continuity, as it organizes memory locally using an entry-point SKILL.md with YAML frontmatter and subfolders for raw, compressed, and state data.
To organize notes for cross-session continuity, store raw handoff notes, compressed summaries, and durable state files under a project-specific directory structure containing separate subfolders for raw, compressed, and state data.
The workflow for recalling project context follows three phases: Phase 1 starts the session, Phase 2 handles mid-session recall of decisions and metrics, and Phase 3 ends the session by generating a compressed memory snapshot from raw notes.
miyamo2Generate spec.json and requirements.md from a project description using templates.
dawsonblockRetrieve relevant long-term memories and return source-aware summaries.
obsfxTrack tasks persistently across sessions using Trekker as the source of truth.
tiankong0101-bytePersist agent context across sessions using session state, vector search, git notes, and markdown archives.
