agent-knowledge CLAUDE.md

Project instructions and technical notes for agent-knowledge, a tool that stores, searches, and displays a codebase’s knowledge in a dashboard. They describe its architecture, code layout, and development rules.

In plain words
What is it for?
Use them when working on the server, dashboard, code-structure extraction, knowledge search, storage, setup, or related hooks and skills.
Why use it?
They give coding agents the project context needed to change the system consistently and safely.

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/keshrath/agent-knowledge/claude-md
Clone the repo
git clone --depth 1 https://github.com/keshrath/agent-knowledge
Per session 1,822 This file is loaded in full into every session.
When invoked 1,822 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 $0.01822 $0.01822
Opus 5 $0.00911 $0.00911
Sonnet 5 $0.00364 $0.00364
Haiku 4.5 $0.00182 $0.00182

Measured 3d ago against content hash f27b18b1472a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-knowledge 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 3d 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 · 140 lines

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.

agent-knowledge

Architecture

Layered architecture — single server.ts handles MCP tools, separate dashboard.ts for HTTP/WebSocket:

scripts/
  tree-sitter-extract.mjs  Standalone CLI for zero-token code structure extraction
  tree-sitter-lang.mjs     Per-language tree-sitter query definitions (8 languages)
  setup.js                 Automated setup wizard (MCP server + hooks + skill install)
  hooks/                   Claude Code lifecycle hooks
skills/
  knowledge-ingest/
    SKILL.md               Claude Code skill file
src/
  server.ts             MCP server, 6 tool definitions, request routing
  dashboard.ts          HTTP + WebSocket server, REST API, file watcher
  index.ts              Entry point (MCP stdio + dashboard auto-start)
  types.ts              KnowledgeConfig, getConfig(), getVersion()
  package-meta.ts       Cached name/version from package.json
  validate.ts           ValidationError class, input validation
  version.ts            Re-exports getVersion from types.ts
  knowledge/
    store.ts            Markdown CRUD, frontmatter parsing, path traversal protection
    search.ts           TF-IDF search over knowledge entries with regex fallback
    git.ts              git pull/push/sync with timeouts
    distill.ts          Session auto-distillation with secrets scrubbing
    graph.ts            Knowledge graph — edges table, link/unlink/traverse (BFS)
    scoring.ts          Confidence/decay scoring — entry_scores table, auto-promotion
    consolidate.ts      Memory consolidation — TF-IDF duplicate detection, cluster grouping
    reflect.ts          Reflection cycle — surfaces unconnected entries, generates prompts
    analyze.ts          Graph analysis — god nodes, bridges, gaps, knowledge brief
  sessions/
    parser.ts           Multi-format session parsing with mtime-based cache
    indexer.ts           Background indexing for sessions
    search.ts           TF-IDF ranked search with 60s global index cache
    scopes.ts           Search scopes (errors, plans, configs, tools, files, decisions)
    summary.ts          Session summaries, topic extraction, file path detection
    adapters/
      index.ts          SessionAdapter interface, adapter registry, auto-init
      opencode.ts       OpenCode adapter (SQLite database)
  search/
    tfidf.ts            TF-IDF scoring engine (tokenizer, stopwords, index)
    fuzzy.ts            Levenshtein distance, sliding window fuzzy matching
    excerpt.ts          Search result excerpt generation
    types.ts            SearchResult, SearchOptions interfaces
  embeddings/
    index.ts            Embedding provider registry
    factory.ts          Provider factory (auto-detect available providers)
    types.ts            EmbeddingProvider interface
    claude.ts           Claude/Voyage embeddings
    openai.ts           OpenAI embeddings
    gemini.ts           Gemini embeddings
    local.ts            Local embedding fallback
  vectorstore/
    index.ts            Vector store facade
    store.ts            SQLite-backed vector storage with cosine similarity
    chunker.ts          Document chunking for embedding
  ui/
    index.html          Dashboard SPA
    styles.css          MD3 design tokens (light + dark)
    app.js              Client-side vanilla JS (WebSocket, tabs, rendering)

Read the full file on GitHub · 140 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. 3d ago First seen · 140 lines · 1,822 tokens per session scan A f27b18b1472a

Subscribe to this mod's changes

agent-knowledge CLAUDE.md is an instructions file published in the GitHub repository keshrath/agent-knowledge (15 stars, last pushed 4mo ago), licensed MIT. It adds 1,822 tokens to every session, about $0.0091 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.

Related

Other instructions, from other repositories