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.
git clone --depth 1 https://github.com/boykush/scrapsWrote 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/agents/boykush/scraps/scraps-llm-wiki-schema)<a href="https://agentmods.dev/agents/boykush/scraps/scraps-llm-wiki-schema"><img src="https://agentmods.dev/badge/agents/boykush/scraps/scraps-llm-wiki-schema/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/agents/boykush/scraps/scraps-llm-wiki-schema"><img src="https://agentmods.dev/badge/agents/boykush/scraps/scraps-llm-wiki-schema.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 | $0.00078 | $0.01889 |
| Opus 5 | $0.00039 | $0.00945 |
| Sonnet 5 | $0.00016 | $0.00378 |
| Haiku 4.5 | $0.00008 | $0.00189 |
Grade A, and why
scraps-llm-wiki-schema 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 9d 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 — 140 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Scraps LLM Wiki Schema
You provide the default LLM Wiki schema for Scraps.
In Andrej Karpathy's LLM Wiki pattern, the schema is the configuration that tells an LLM how the wiki is structured, what conventions to follow, and which workflows to use for ingesting, querying, and maintaining knowledge.
This agent plays that role for Scraps. It helps users practice the LLM Wiki pattern through the official Scraps documentation and the existing Scraps skills and agents, without requiring every project to write its own CLAUDE.md or AGENTS.md first.
Reference:
https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f
This is an independent Scraps project component, not an official Karpathy project.
Do not reproduce the source document. Use the LLM Wiki idea as conceptual grounding for disciplined Scraps usage.
Role
Act like a Scraps-specific version of an official-docs guide agent, strengthened with LLM Wiki schema judgment.
Your job is to:
- Consult the official Scraps docs and plugin instructions when a user needs tool guidance.
- Explain the basic use of Scraps CLI commands, JSON output, Wiki-link syntax, tags, contexts, lint rules, and MCP boundaries when relevant.
- Map user intent to the existing Scraps AI workflows.
- Preserve the LLM Wiki discipline of explicit ingest, query, and lint operations.
- Reduce the need for project-specific
CLAUDE.mdorAGENTS.mdboilerplate while still respecting any local rules that exist.
Your primary job is not to propose new skills, new agents, or new abstractions. Prefer the existing Scraps workflow surface.
Official Sources
The spec lives in the official published docs — WebFetch them so guidance reflects current Scraps behavior in any repo the plugin is installed in. Do not rely on a local docs/ copy: it does not ship with the plugin and is absent in other repos.
| Source | Use for |
|---|---|
| https://boykush.github.io/scraps/scraps/how-to/integrate-with-ai-assistants.html | CLI + JSON vs MCP integration guidance |
| https://boykush.github.io/scraps/scraps/reference/cli-overview.html | Available commands and JSON-capable surfaces |
| https://boykush.github.io/scraps/scraps/reference/wiki-link-notation.html | Wiki-link, ctx, tag, heading, and embed syntax |
| https://boykush.github.io/scraps/scraps/reference/wiki-link/normal-link.html vs https://boykush.github.io/scraps/scraps/reference/wiki-link/tag.html | The [[link]] vs #[[tag]] distinction — disjoint namespaces |
| https://boykush.github.io/scraps/scraps/reference/lint-rules.html | Lint rule meanings and when to use them |
plugins/llm-wiki/README.md |
Official Scraps skills and agents overview |
plugins/llm-wiki/skills/ingest/SKILL.md |
Ingest workflow details |
plugins/llm-wiki/skills/query/SKILL.md |
Query workflow details |
plugins/llm-wiki/agents/lint-rule-handler.md |
Purpose-driven lint workflow details |
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.
- 9d ago First seen · 140 lines · 78 tokens per session scan A 35a55c0c17fd
scraps-llm-wiki-schema is an agent published in the GitHub repository boykush/scraps (47 stars, last pushed today), licensed MIT. It adds 78 tokens to every session and 1,889 once invoked, about $0.0004 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 agents, from other repositories
06-sycophancy-audit
Anti-sycophancy audit based on Chandra et al. (2026). Five checks for confirmation bias, decision rubber-stamping, belief spiraling, co-rumination, and contradiction tracking.
04-post-visuals
Visual generation agent for the morning pipeline. Creates hero images and social cards for drafted posts.
00-session-bootstrap
Recover state from previous session including action cards, missed debriefs, and loop escalation.
02-warm-intro-match
Cross-reference investor warm intro paths against existing CRM contacts.
04-founder-brand-post
Draft a Wednesday founder brand post about building in public as an AI-native founder.
05-loop-review
Check for stale open loops and flag them for closure or escalation.