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 instructions/awrshift/agent-memory-kit/claude-mdgit clone --depth 1 https://github.com/awrshift/agent-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 | $0.00594 | $0.00594 |
| Opus 5 | $0.00297 | $0.00297 |
| Sonnet 5 | $0.00119 | $0.00119 |
| Haiku 4.5 | $0.00059 | $0.00059 |
Grade A, and why
agent-memory-kit CLAUDE.md 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 yesterday.
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.
This is a copy
100% identical to claude-memory-kit CLAUDE.md — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 45 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Working on Memory Kit
This repo is a Claude Code plugin marketplace, not a memory workspace. The plugin lives in
plugins/memory-kit/; .claude-plugin/marketplace.json is the catalog. Users install it into
their repositories — nothing here should assume it runs at the project root.
Rules that are easy to get wrong
- Paths. Anything the plugin ships resolves from
${CLAUDE_PLUGIN_ROOT}; anything about the user's repo resolves fromCLAUDE_PROJECT_DIR. Never derive a project root from__file__— the plugin runs from a cache directory. - Never write into a repo that didn't ask. The plugin can be installed user-wide. Scaffolding
happens only inside
/memory-kit:setup, after a yes. - What a plugin cannot ship:
CLAUDE.md,.claude/rules/, and anysettings.jsonkey other thanagent/subagentStatusLine. Context therefore travels via hooks, skills and agents — the working agreement isplugins/memory-kit/context/identity.md, injected at SessionStart. - Always-loaded vs on-invoke. A skill body loads only when invoked; a rule without
paths:loads every session forever. Put weight in skills andreference/, never in rules. - Verify injection by looking at the context, not the code. v5 claimed for a year that the hot cache was always loaded while the hook only measured it. Run the hook and read its output.
Before committing a change to hooks or skills
claude plugin validate ./plugins/memory-kit
CLAUDE_PROJECT_DIR=<test-repo> CLAUDE_PLUGIN_ROOT=$PWD/plugins/memory-kit \
python3 plugins/memory-kit/hooks/session-start.py <<< '{"source":"startup"}'
Exercise every SessionStart source (startup · resume · compact · fork · clear) and both hook
outcomes for protect-tests.py — the profiles are the whole design, and a broken one fails
silently.
Docs that must stay true in the same commit
A behaviour change touches docs/CHANGELOG.md; a layer change also touches docs/ARCHITECTURE.md
and BOTH readmes — the root README.md (the pitch) and plugins/memory-kit/README.md (the skill
table). A change to what a diagram asserts touches docs/ASSETS.md, even when the PNG is not
regenerated in the same commit — a stale panel is a lie in the most-read file. VERSION,
plugin.json and marketplace.json carry the same version — they have drifted before.
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.
- yesterday First seen · 45 lines · 594 tokens per session scan A 1b0ea90728d9
agent-memory-kit CLAUDE.md is an instructions file published in the GitHub repository awrshift/agent-memory-kit (31 stars, last pushed 2d ago), licensed MIT. It adds 594 tokens to every session, about $0.0030 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to claude-memory-kit CLAUDE.md, differing in 0 lines, and is treated as a copy.
Other instructions, from other repositories
kungfu AGENTS.md
AGENTS.md instructions for kungfu-systems/kungfu, covering agents.md, start from the person's objective, rules that always apply, load bounded context before changing the repository and execute and verify through shifu.
kungfu CLAUDE.md
Claude Code instructions for kungfu-systems/kungfu: Read and follow AGENTS.md before working in this repository. It routes both product use and contribution work to the repository sources of truth, including the requirement that development and build tasks enter through Shifu.
kungfu copilot-instructions.md
Copilot instructions for kungfu-systems/kungfu: Read and follow AGENTS.md before proposing or running changes. It is the shared agent router for this repository, including the canonical Shifu development and build entrypoint.
mnemon AGENTS.md
Instructions for mnemon-dev/mnemon, covering mnemon agent guidelines, development, go engineering and commit discipline.
amfs CLAUDE.md
Instructions for raia-live/amfs, covering amfs memory — agent instructions, available mcp tools, identity, brain tools (agent-scoped) and shared knowledge tools.
dotagents AGENTS.md
Instructions for bgreenwell/dotagents, covering agents.md, project purpose, context routing and validation.