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 skills add anh-chu/llm-wiki-pm --skill llm-wiki-maintaingit clone --depth 1 https://github.com/anh-chu/llm-wiki-pmWrote 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/skills/anh-chu/llm-wiki-pm/llm-wiki-maintain)<a href="https://agentmods.dev/skills/anh-chu/llm-wiki-pm/llm-wiki-maintain"><img src="https://agentmods.dev/badge/skills/anh-chu/llm-wiki-pm/llm-wiki-maintain/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/anh-chu/llm-wiki-pm/llm-wiki-maintain"><img src="https://agentmods.dev/badge/skills/anh-chu/llm-wiki-pm/llm-wiki-maintain.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00050 | $0.01256 |
| Opus 5 | $0.00025 | $0.00628 |
| Sonnet 5 | $0.00010 | $0.00251 |
| Haiku 4.5 | $0.00005 | $0.00126 |
Grade A, and why
llm-wiki-maintain 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 11d 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 — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
LLM Wiki Maintain
The recurring maintenance loop, as one defined operation. "Every day: sweep all sources → ingest what's new → produce a brief → keep the wiki tidy" was previously improvised step-by-step on faith; this skill makes the order, the brief lifecycle, and the autonomous-mode behavior explicit.
This skill orchestrates — it does not reimplement. The actual work is done
by the core llm-wiki-pm skill (orient, ingest, lint) and llm-wiki-brief
(brief generation). Route to them at each step. If a sub-skill isn't installed,
fall back to the core skill's matching operation.
Mode: interactive vs autonomous (decide first)
The skill's proactive behaviors (ambient "want me to add that?", per-step confirmations, post-task offers) assume a human is watching. In a scheduled or autonomous run nobody answers them — they stall or force a guess. So classify the run before doing anything:
- Interactive — a human invoked this in a live chat. Normal behavior: surface takeaways, ask before mass updates (10+ pages), offer follow-ups.
- Autonomous — a scheduled/cron/loop run with no human in the loop (no one
to answer a prompt). Then:
- Suppress interactive prompts. Do not ask "want me to add this?" or wait on confirmations. Act on the clear cases.
- Act-then-report, never block. Anything that would need a human decision (a risky mass update, an ambiguous create-vs-update, a possible contradiction) goes into a Needs Review queue in the brief instead of halting the run.
- Stay within safe bounds. Never delete or supersede without a human; queue those. Snapshot before any overwrite (core rule still applies).
If you can't tell which mode you're in, ask once ("interactive or autonomous run?") and proceed on the answer; default to interactive.
The loop (ordered)
① Orient. Run the core Pre-Flight + orient (SCHEMA, index, log, overview,
_status.md). Do not skip — every write still requires orient.
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.
- 11d ago First seen · 91 lines · 50 tokens per session scan A 1addfd31ddd0
llm-wiki-maintain is a skill published in the GitHub repository anh-chu/llm-wiki-pm (6 stars, last pushed 1mo ago), licensed MIT. It adds 50 tokens to every session and 1,256 once invoked, about $0.0003 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
llm-wiki
Build and maintain a persistent, interlinked Obsidian-compatible markdown wiki using Karpathy's LLM Wiki pattern. Extension-backed with auto-generated metadata, guardrails, and 14 custom tools (+3 opt-in agent-trajectory tools).
mindbase
Maintain a persistent AI-tended research wiki. Activates when the user wants to ingest a source (PDF, URL, article, paste), capture a thought, build the wiki context, query what's already known, lint for orphans/contradictions, or maintain page conventions. Phrases include "add this to my wiki", "today I decided…"…
wiki-crystallize
Distil a chat thread, research session, or working document into a structured wiki page capturing the current state of knowledge. Always use this skill when the user says /wiki-crystallize, 'save this to my wiki', 'capture what we've worked out', 'write this up as a wiki page', 'update my wiki from this conversation'…
wiki-query
Answer a question using the compiled wiki knowledge base, synthesising a response with [[wikilink]] citations. Always use this skill when the user says /wiki-query, 'what does my wiki say about', 'what do I know about', 'check my notes on', 'search my wiki for', or 'what's the current state of [topic] in my notes'.…
wiki-fold
Roll up the wiki's journal.md entries into structured fold pages — like 2^k log compaction. Reads the last 2, 4, 8, 16... entries and writes a fold page that summarizes them by extractive summarization (no invention), with backlinks to children. Idempotent at the structural level — re-running with the same window…
wiki
A persistent Markdown knowledge base maintained by an AI agent. Markdown is a plain-text format for structured notes; the knowledge base can hold personal research or documentation about an open-source project's design.