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 instructions/lling0000/siftmarks/claude-mdgit clone --depth 1 https://github.com/Lling0000/SiftMarksWrote 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/instructions/lling0000/siftmarks/claude-md)<a href="https://agentmods.dev/instructions/lling0000/siftmarks/claude-md"><img src="https://agentmods.dev/badge/instructions/lling0000/siftmarks/claude-md.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 | $0.02337 | $0.02337 |
| Opus 5 | $0.01169 | $0.01169 |
| Sonnet 5 | $0.00467 | $0.00467 |
| Haiku 4.5 | $0.00234 | $0.00234 |
Grade A, and why
SiftMarks 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.
How it starts
The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Commands
- Install:
npm install - Build TS packages + CLI + MCP server (skips Next build):
npm run build:packages - Build everything (packages + web):
npm run build - Run the local dashboard at
http://localhost:4399:npm run dev - Typecheck the workspace via project references:
npm run typecheck(buildstsconfig.jsonreferences; the web app is not listed, so this passes even when Next would fail — see Caveats) - Lint:
npm run lint(root config is minimal; the web app uses its owneslint.config.mjsvianpm run lint -w apps/web) - Run the CLI after
build:packages:npm run cli -- <command>(e.g.init,stats,doctor,search "...",rescue,index --limit 100,import ./bookmarks.html,export out.json,mcp) - Start MCP server directly:
node apps/mcp-server/dist/index.js - There is no test runner configured — no
npm test, no test files. Don't claim tests pass; verify by running CLI commands or hitting API routes.
Per-workspace builds use tsc (or next build for apps/web); workspaces are wired via npm workspaces. Build order matters: shared → db → ai → core → indexer → cli → mcp-server → web. npm run build:packages enforces this order — if you build a single workspace, build its deps first.
For experiments that should not touch the user's real bookmark library at ~/.siftmarks/, set SIFTMARKS_HOME=/tmp/siftmarks-test (read by getDataDir() in packages/db/src/database.ts).
Architecture
SiftMarks is a local-first TypeScript monorepo that turns browser bookmarks into a searchable, AI-augmented memory library. Everything runs on the user's machine; no account, no telemetry. The app data lives in a single SQLite file at ~/.siftmarks/siftmarks.sqlite (overridable via SIFTMARKS_HOME).
The four entry points share one database
All four apps open the same SQLite file via new SiftMarksDB() and call .initialize() on startup. database.ts is the single source of truth for SQL — neither the web routes nor the MCP tools issue raw SQL. Schema (packages/db/src/schema.ts) is idempotent (CREATE TABLE IF NOT EXISTS), so any entry point can boot a fresh DB.
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.
- 3d ago First seen · 99 lines · 2,337 tokens per session scan A 14916fddaa0e
SiftMarks CLAUDE.md is an instructions file published in the GitHub repository Lling0000/SiftMarks (24 stars, last pushed 2mo ago), licensed MIT. It adds 2,337 tokens to every session, about $0.0117 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 instructions, from other repositories
hindsight CLAUDE.md
Claude Code instructions for vectorize-io/hindsight, covering claude.md, project overview, development commands, local development (api + ui) and start both api server and control plane ui.
hindsight AGENTS.md
AGENTS.md instructions for vectorize-io/hindsight: See CLAUDE.md for project documentation and coding conventions.
mulch CLAUDE.md
Instructions for jayminwest/mulch, covering claude.md, build & test commands, architecture, storage model and record types & classifications.
mulch AGENTS.md
Instructions for jayminwest/mulch, covering agents.md, what this project is, tech stack at a glance, project layout and commands.
karakeep CLAUDE.md
Claude Code instructions for karakeep-app/karakeep, a project described as: A self-hostable bookmark-everything app (links, notes and images) with AI-based automatic tagging and full text search.
agentor copilot-instructions.md
Instructions for CelestoAI/agentor: Read AGENTS.md in the repository root. It is the single source of contributor guidance for this project — architecture, development setup, code style, testing, and CI — and applies to Copilot exactly as written.