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/krishddd/llm-wiki/agents-mdgit clone --depth 1 https://github.com/krishddd/llm-wikiWhat 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 | $0.01366 | $0.01366 |
| Opus 5 | $0.00683 | $0.00683 |
| Sonnet 5 | $0.00273 | $0.00273 |
| Haiku 4.5 | $0.00137 | $0.00137 |
Grade A, and why
llm-wiki 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 — 106 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — LLM-Wiki agent surface
Agent-facing companion to
CLAUDE.md. Lists the tools, resources, and conventions an MCP-aware agent (Claude Code, Cursor, Codex, etc.) must follow when driving this wiki. Loaded byllm_wiki/mcp_server.py.
Quick start for an agent
- Call
context_start(days=7, top_pages=5)at the beginning of every new session. Returns recent episodic timeline + top accessed pages + open contradictions. - To answer a user question: call
query_wiki(question, top_k=5, save_back=true). The response carries numbered citations, per-claim confidences, and follow-up suggestions. - After a chain of related queries, call
session_crystallize(correlation_ids=[...])to persist the exploration as one durable wiki page. - To add a document: call
ingest_file(path)with an absolute path on the host. - Periodically (or when results feel stale) call
run_admin_job("lint_autofix")orrun_admin_job("decay_sweep")— same code paths as the in-process scheduler.
Tool catalogue (MCP)
| Tool | Purpose |
|---|---|
query_wiki(question, top_k=5, save_back=true) |
Hybrid retrieval + reflective synthesis. Returns answer with [1] [2]-style citations, blocks, follow-ups, intent, retrieval quality. |
ingest_file(path) |
Multi-format ingest (PDF/DOCX/HTML/PPTX/XLSX/CSV/MD/TXT). Confidence-gated write. |
lint_wiki() |
Read-only health report. |
list_entities(limit=50) |
Canonical entities by backlink count. |
session_crystallize(correlation_ids, days=14) |
Distil an exploration thread into one wiki page. |
context_start(days=7, top_pages=5) |
Session-start briefing. |
run_admin_job(job_name) |
Manually trigger one of: decay_sweep, episodic_prune, promote_episodic, review_autopilot, lint_autofix, detect_procedures, build_topics, page_compaction, lint. |
Resources (MCP)
| Resource URI | Content |
|---|---|
wiki://index |
The current wiki/index.md (master catalogue, regenerated deterministically per ingest). |
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 · 106 lines · 1,366 tokens per session scan A 765d59a39c6f
llm-wiki AGENTS.md is an instructions file published in the GitHub repository krishddd/llm-wiki (2 stars, last pushed 10d ago), licensed MIT. It adds 1,366 tokens to every session, about $0.0068 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 instructions, from other repositories
agency-orchestrator CLAUDE.md
Instructions for jnMetaCode/agency-orchestrator, covering agency orchestrator — project context, key commands, community templates (remote manifest), release pipeline (all automatic) and skills (methodology playbooks).
PlanExe AGENTS.md
Instructions for PlanExeOrg/PlanExe, covering planexe agent instructions, repo architecture map, shared contracts, hard rules (agent safety) and environment variables.
Elpis AGENTS.md
Instructions for MasihMoafi/Elpis, covering elpis agent map, start, context discipline, definition of done and evals first.
loomcycle CLAUDE.md
Instructions for denn-gubsky/loomcycle, covering claude.md — loomcycle project guide, project context, development workflow, karpathy guidelines (coding discipline) and security rules — non-negotiable.
zeph AGENTS.md
Instructions for bug-ops/zeph, covering repository guidelines, project structure & module organization, specifications (mandatory), agent operating notes and multi-model design.
nodetool AGENTS.md
AGENTS.md instructions for nodetool-ai/nodetool, covering nodetool — agent rules, communication style, instructions, 1. positive patterns and negative patterns and 2. reference points.