Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/lh8ppl/core-memory-kit/memory-writenpx skills add LH8PPL/core-memory-kit --skill memory-writegit clone --depth 1 https://github.com/LH8PPL/core-memory-kitWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00202 | $0.01350 |
| Opus 5 | $0.00101 | $0.00675 |
| Sonnet 5 | $0.00040 | $0.00270 |
| Haiku 4.5 | $0.00020 | $0.00135 |
Grade A, and why
memory-write scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 130 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Capturing durable memory
Durable facts — preferences, decisions, environment state — are saved through the kit's safe write path (Poison_Guard secret screening + home-path sanitization + dedup + conflict detection).
- NEVER hand-edit
context/MEMORY.md,context/USER.md, or any file undercontext/memory/. Direct edits bypass screening and can leak a credential or a local path into a committed file. - Silent on success. Do not announce "saved to memory" unless the user asked.
There are two equivalent surfaces onto the same safe path. Prefer the MCP tools
when the cmk server is connected — params are structured data, so backtick /
$() / quote-heavy rationale can't be mangled by a shell, and there's no
per-command approval prompt.
Preferred: the cmk MCP tools (when connected)
- Capture → call
mk_rememberwithtext. For a preference, working-style rule, or constraint, also passwhy,how,title, andtype— this writes a rich Why/How fact file, not just a bullet. - Remove → call
mk_forgetwith the factid. Two-step: the first call previews what would be removed and returns aconfirm_token; call again with that token to tombstone (audit trail preserved). Confirm with the user first. - Adjust trust → call
mk_trustwith the factidand aleveloflow,medium, orhigh. Use when the user signals how much a saved fact matters: "trust this" / "that's important — keep it" →high; "that's not important / I'm not sure / low priority" →low. Trust drives what gets injected first and what ages out, so this is the user steering their own memory without editing files.
type is one of:
feedback— how the user wants you to workuser— who the user is (role, expertise)project— an ongoing goal or constraintreference— a pointer to an external resource (URL, ticket, dashboard)
Fallback: the cmk CLI (when the MCP server isn't connected)
Capture a bullet:
cmk remember "<the fact, one sentence>"
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 130 lines · 202 tokens per session scan A 6374a26c2830
memory-write is a skill published in the GitHub repository LH8PPL/core-memory-kit (6 stars, last pushed 5d ago), licensed MIT. It adds 202 tokens to every session and 1,350 once invoked, about $0.0010 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
Effective Memory
The essential habits for an AI agent with memory — session bookends, learning triggers, verification, safety, and the operational discipline that turns raw recall into compounding intelligence. Pinned, always-injected.
recall
Recall this repository's OwnMem local memory before changing code, and keep it healthy. Use when a repository contains .ownmem/, when past debugging lessons could apply ("have we hit this before", "why is it done this way"), or when the user mentions ownmem, project memory, or recalling across sessions.
mnemo-cortex
Installs and wires Mnemo Cortex (local-first persistent memory) into OpenClaw and other MCP-capable agents. Use for cross-session recall, decision history, or multi-agent shared memory.
init
Install or update OwnMem in the current repository. Use when the user asks to set up OwnMem, add local project memory for coding agents, or refresh an existing OwnMem installation after a version bump.
ori-memory
Persistent agent memory with learning retrieval. Knowledge graph on markdown files — capture insights, decisions, research, and learnings during work, then retrieve them weeks or months later. Use when knowledge is too valuable to lose but too much to inject into every prompt.
ogham-research
Structured memory capture for Ogham shared memory. Use when the user wants to store findings, remember something, save what was learned, or capture a decision. Triggers on "remember this", "store this", "save this finding", "save what we learned", "capture this decision", "log this", or any request to persist…