daedra CLAUDE.md

daedra CLAUDE.md is an instructions file for coding agents from dirmacs/daedra. It costs 695 tokens per session, scanned A, original, MIT.

A self-contained MCP server for searching the web, opening pages, and crawling websites. It combines several search sources and can extract readable text from web pages, PDFs, and office documents.

In plain words
What is it for?
Use it to search the web, visit a specific page, or crawl a site. It can handle HTML, PDFs, office files, ebooks, text, SVG, and JSON, while rejecting unsupported binary files.
Why use it?
It provides a single search and page-reading interface when one search source is unavailable or limited. It also turns supported documents into text that an AI assistant can inspect.

Instructions file

Install

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.

agentmods
npx agentmods add instructions/dirmacs/daedra/claude-md
Clone the repo
git clone --depth 1 https://github.com/dirmacs/daedra

Wrote 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.

agentmods badge for daedra CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/dirmacs/daedra/claude-md.svg)](https://agentmods.dev/instructions/dirmacs/daedra/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/dirmacs/daedra/claude-md"><img src="https://agentmods.dev/badge/instructions/dirmacs/daedra/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 695 This file is loaded in full into every session.
When invoked 695 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00695 $0.00695
Opus 5 $0.00347 $0.00347
Sonnet 5 $0.00139 $0.00139
Haiku 4.5 $0.00069 $0.00069

Measured 4d ago against content hash dd2f1e48c71b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

daedra CLAUDE.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 4d 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.

CLAUDE.md · 44 lines

How it starts

The opening of the file, as written. The whole thing — 44 lines — stays where its author put it; the contents beside it link to each section on GitHub.

daedra

Self-contained web search MCP server. 14 unkeyed backends with automatic fallback. Mwmbl and Marginalia answer from any IP. Brave HTML may rate-limit. No paid search keys.

Build and test

cargo build --release
cargo test
cargo clippy -- -D warnings

Architecture

Single crate with modular backends: Mwmbl, Brave, Marginalia, Bing RSS, Bing, Google News, Hacker News, Google, Wikipedia, StackExchange, GitHub, Wiby, DDG Instant, DuckDuckGo HTML. SearchProvider runs the fallback chain. Per-backend circuit breakers (BackendHealth), governor keyed rate limits, and classified retry (transient errors only) keep the chain stable.

FetchClient classifies responses as FetchedContent (Html / Pdf / Document / Text / Binary). HTML uses dom_smoothie Readability extraction. PDFs use infer + pdf-inspector (Markdown out). Office/ebook documents use anydoc. Textual SVG/XML/JSON come back verbatim. Zip containers typed as application/zip are sniffed for embedded office formats. Binary types fail with a typed error.

The MCP server (DaedraHandler in server.rs) exposes web_search, visit_page, and crawl_site. search_duckduckgo is a backward-compatibility alias for web_search. Transports: STDIO and SSE (Axum). The moka async cache layer caches results.

Key files

  • src/main.rs — CLI entrypoint (Commands::run, CheckReporter for health checks)
  • src/server.rs — MCP server, Axum HTTP/SSE, tool handler methods
  • src/lib.rs — Crate root and re-exports
  • src/url_classification.rs — Search result URL → content type (data-driven rules)
  • src/tools/backend.rs — SearchProvider, circuit breakers, rate limiters, fallback chain, aggregate failure messages
  • src/tools/rss.rs — Machine-format backends: Bing RSS, Google News, HN Algolia
  • src/tools/soft_block.rs — Zero-result page classifier for scraper backends
  • src/tools/fetch.rs — Page fetch, Readability, PDF, MIME classification
  • src/tools/crawl.rs — Multi-page crawl tool
  • src/tools/ — Individual backend implementations
  • src/cache.rs — moka async cache layer
  • src/types.rs — Shared schemas. Language and topic detection tables.

Read the full file on GitHub · 44 lines

Changes

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.

  1. 4d ago First seen · 44 lines · 695 tokens per session scan A dd2f1e48c71b

Subscribe to this mod's changes

daedra CLAUDE.md is an instructions file published in the GitHub repository dirmacs/daedra (13 stars, last pushed yesterday), licensed MIT. It adds 695 tokens to every session, about $0.0035 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.