LLM Wiki is a system that uses AI agents to research sources, compile knowledge bases, organize project briefs, and answer questions from the resulting wiki. It is used to turn rough ideas and investigations into structured, reusable knowledge for agents, including Claude Code, Codex, and OpenCode. The catalogue entries are commands, hooks, skills, instructions, and a plugin that implement its workflow.
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 commands/nvk/llm-wiki/specialistgit clone --depth 1 https://github.com/nvk/llm-wikiWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/nvk/llm-wiki/specialist)<a href="https://agentmods.dev/commands/nvk/llm-wiki/specialist"><img src="https://agentmods.dev/badge/commands/nvk/llm-wiki/specialist.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.00020 | $0.00849 |
| Opus 5 | $0.00010 | $0.00425 |
| Sonnet 5 | $0.00004 | $0.00170 |
| Haiku 4.5 | $0.00002 | $0.00085 |
Grade A, and why
specialist 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 5d 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 — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Your task
Manage or apply personal specialist methods. Read
skills/wiki-manager/references/specialists.md before acting. A specialist is
a bounded evidence/review protocol, not a human credential or new authority.
Resolve HUB from $HOME/.config/llm-wiki/config.json, preferring hub_path and
expanding only a leading ~. Use the bundled deterministic helper:
"${CLAUDE_PLUGIN_ROOT}/bin/llm-wiki" specialist <subcommand>
Use scripts/llm-wiki in a source checkout. Never assume a global install.
Deterministic subcommands
Route init, create, refresh, list, show, validate, enable, and
disable directly to the helper with the user's arguments. For create, help
the user choose a descriptive method name rather than a credential costume,
then review and replace the scaffold's generic instructions and TODOs. Run
validate, then refresh; do not enable an unfinished scaffold.
enable and disable require an active hub topic. V1 does not persist
allowlists inside project-local .wiki/ roots. Never edit registry.json
manually when the helper can make the change.
suggest
Discover useful specialist candidates without creating them:
- Read
HUB/_index.md, activewikis.jsonentries, and.skills/_index.mdwhen present. - Read each selected active topic's root
_index.md. Do not bulk-read raw sources or recursively scan the hub. - Rank recurring decision patterns and evidence-review needs across topics. Use targeted category indexes and a small content sample only to verify the strongest patterns.
- Return at most ten candidates. For each include: proposed method name, topics/use cases that justify it, mandate, exclusions, risk tier, source hierarchy, and two starter eval cases.
- Separate broad reusable methods from narrow topic-only methods. Prefer the former only when recurrence is demonstrated.
- Do not create or enable any candidate without a later explicit request.
If the user asks to save the result, write a dated report under the selected
topic's output/, update output/_index.md and the topic root _index.md, and
append log.md.
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.
- 5d ago First seen · 75 lines · 20 tokens per session scan A fda58b3a4996
specialist is a command published in the GitHub repository nvk/llm-wiki (1,188 stars, last pushed 8d ago), licensed MIT. It adds 20 tokens to every session and 849 once invoked, about $0.0001 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 commands, from other repositories
gen-plan
Generate implementation plan from draft document.
refine-plan
Refine an annotated implementation plan and generate a QA ledger.
start-rlcr-loop
Start iterative loop with Codex review.
gen-idea
Generate a repo-grounded idea draft via directed-swarm exploration.
sync_roleplay_docs
当用户调用此命令时，请同步整理“角色扮演模式（个人扮演模式）”相关的规则、技能、spec 和测试收尾项。.
sync_contributors
当用户调用此命令时，请同步并更新仓库根目录 CONTRIBUTORS.md 以及 README.md 中的贡献者部分，来源为 GitHub 仓库贡献者列表。.