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/ctxr-dev/llm-wiki-memory/agents-mdgit clone --depth 1 https://github.com/ctxr-dev/llm-wiki-memoryWhat 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 | $0.02138 | $0.02138 |
| Opus 5 | $0.01069 | $0.01069 |
| Sonnet 5 | $0.00428 | $0.00428 |
| Haiku 4.5 | $0.00214 | $0.00214 |
Grade A, and why
llm-wiki-memory AGENTS.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 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 — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
llm-wiki-memory: local LLM-wiki-backed memory for AI coding agents. Capture, compile,
recall, stored as leaves in a git-versioned local wiki (via @ctxr/skill-llm-wiki), with
local-embedding recall. No RAG, no Docker.
Development discipline (.agents/)
Rules for working ON this repo are canonical in .agents/rules/:
live-runtime-safety.md (read first — this tree is executed live by the MCP server,
hooks and cron; editing it can destroy live data),
dev-principles.md (durability / parsing / injection / config invariants, cross-client
portability, wiki placement, LLM-pipeline contracts, hook design),
defensive-invariants.md (when a guard may be deleted; guard vs tripwire; why
"unreachable by derivation" is not "unreachable in fact"),
verification-completeness.md (what "green" is allowed to mean — the full gate chain,
falsifiable tests, named before/after invariants for durable artifacts),
module-state-ownership.md (one state machine per module; keyed memos over hot-reloading
settings; the disposal duty a rebuild path creates),
testing.md (harness + mocking conventions, the /tmp/lwm-* leak trap),
release-migrations-authoring.md (a breaking change ships an idempotent,
state-detecting migration under scripts/migrations/ — never a runbook),
self-observability.md (capture confirmed engine bugs while you work),
and docs-style.md (README/docs conventions). Step-by-step procedures live in
.agents/skills/ (write-migration.md, run-tests-safely.md,
debug-capture-pipeline.md, verify-live-change.md, triage-review-findings.md).
Per-client shadows reference the canonical files via @-imports — .claude/rules/ +
.claude/skills/ (Claude Code), .cursor/rules/ (Cursor); always edit the .agents/
file, never a shadow. The shadows are GENERATED: run npm run wire:dev-surfaces after
adding or renaming a canonical file, and test/dev-surface-mirror.test.mjs fails the
gate if any client is missing one. That gate exists because Cursor silently went five
shadows short — a missing shadow means the rule simply does not apply in that client,
and nothing reports it.
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 · 120 lines · 2,138 tokens per session scan A dc77ef81f2f5
llm-wiki-memory AGENTS.md is an instructions file published in the GitHub repository ctxr-dev/llm-wiki-memory (129 stars, last pushed 5d ago), licensed MIT. It adds 2,138 tokens to every session, about $0.0107 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 2026-08-30.
Other instructions, from other repositories
dense-mem AGENTS.md
Instructions for markhuangai/dense-mem, covering dense-mem repository guidance, project context, architecture decision records, current stack and target architecture.
honcho CLAUDE.md
Claude Code instructions for plastic-labs/honcho, covering claude.md, honcho overview, what is honcho?, core concepts and peer paradigm.
altk-evolve AGENTS.md
Instructions for AgentToolkit/altk-evolve, covering what is evolve?, key concepts, architecture flow, project directory tree (some files omitted for brevity) and first time setup.
AIPass AGENTS.md
Instructions for AIOSAI/AIPass, covering aipass, startup protocol and memories.
caura-cross-fleet-gov CLAUDE.md
Instructions for caura-ai/caura-cross-fleet-gov, covering claude.md, what this repo is, prerequisites, environment setup and gateway commands.
mfs CLAUDE.md
Instructions for zilliztech/mfs, covering claude.md, what this is, commands, architecture that spans files and connector config: four sources, one source of truth.