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/basalt-ai/cobalt/claude-mdgit clone --depth 1 https://github.com/basalt-ai/cobaltWrote 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/basalt-ai/cobalt/claude-md)<a href="https://agentmods.dev/instructions/basalt-ai/cobalt/claude-md"><img src="https://agentmods.dev/badge/instructions/basalt-ai/cobalt/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 | $0.03875 | $0.03875 |
| Opus 5 | $0.01937 | $0.01937 |
| Sonnet 5 | $0.00775 | $0.00775 |
| Haiku 4.5 | $0.00387 | $0.00387 |
Grade A, and why
cobalt 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 — 545 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — Development Guide
💡 Collaboration First: This project values collaborative decision-making. When facing decisions about architecture, developer experience, or anything with significant impact, always present options and wait for user input before proceeding. See Decision-Making Philosophy for details.
Project Overview
Cobalt is a TypeScript CLI testing framework for AI agents and LLM-powered applications. Think "Unit testing for AI Agents" — it provides experiment runners, evaluators, datasets, and result tracking.
Tech Stack
- Language: TypeScript 5.7 with strict mode
- Build Tool: tsup for bundling
- CLI Framework: citty for command-line interface
- Testing: Vitest for unit and integration tests
- Code Quality: Biome for linting and formatting
- Database: better-sqlite3 for history tracking
- HTTP Server: Hono for dashboard API
- Package Manager: pnpm
Architecture
Cobalt follows a clean, modular architecture:
packages/cobalt/
├── src/
│ ├── core/ # Core experiment runner logic
│ │ ├── experiment.ts # Main experiment() function
│ │ ├── Evaluator.ts # Evaluator class
│ │ └── config.ts # Configuration system
│ ├── datasets/ # Dataset loading and transformation
│ │ └── Dataset.ts # Dataset class
│ ├── evaluators/ # Evaluator implementations
│ │ ├── llm-judge.ts # LLM-based evaluation
│ │ ├── function.ts # Custom function evaluation
│ │ ├── exact-match.ts # String matching
│ │ └── similarity.ts # Embeddings (P2 - not implemented)
│ ├── cli/ # CLI commands
│ │ ├── index.ts # CLI entry point
│ │ └── commands/ # Individual commands (run, init, etc.)
│ ├── dashboard/ # Dashboard server (P4)
│ │ ├── server.ts # Hono server
│ │ └── routes.ts # API routes
│ ├── mcp/ # Model Context Protocol (P3)
│ │ ├── server.ts # MCP server
│ │ └── tools.ts # MCP tool implementations
│ ├── storage/ # Data persistence
│ │ ├── results.ts # JSON result files
│ │ ├── cache.ts # LLM response cache
│ │ └── db.ts # SQLite history database
│ ├── utils/ # Utilities
│ │ ├── cost.ts # Token cost estimation
│ │ ├── stats.ts # Statistical calculations
│ │ ├── template.ts # Template rendering
│ │ └── hash.ts # Hash generation
│ └── types/ # TypeScript types
└── tests/ # Test suite
├── unit/ # Unit tests
├── integration/ # Integration tests
└── helpers/ # Test helpers and mocks
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 · 545 lines · 3,875 tokens per session scan A b88e930e46c0
cobalt CLAUDE.md is an instructions file published in the GitHub repository basalt-ai/cobalt (68 stars, last pushed 6mo ago), licensed MIT. It adds 3,875 tokens to every session, about $0.0194 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 2026-09-01.
Other instructions, from other repositories
GPT-RAG release.instructions.md
Instructions for Azure/GPT-RAG, a project described as: Enterprise-grade accelerator for agentic RAG on Azure. Built on Microsoft Foundry with Foundry IQ as the default retrieval backend, Microsoft Agent Framework orchestration, Zero-Trust architecture and IaC.
apex-accelerator vendor-prompting.instructions.md
Vendor prompting best-practice rules for Anthropic Claude and OpenAI GPT-5.6-Terra agents and prompts. Each rule cites a rule ID in the vendor-prompting skill rules.json registry. Validator: npm run lint:vendor-prompting.
ken CLAUDE.md
Claude Code instructions for townsendmerino/ken, covering claude.md, what this is, repository ownership (read this first), commands and embedding parity & golden fixtures (now in aikit).
rag-code-mcp copilot-instructions.md
Instructions for doITmagic/rag-code-mcp, covering copilot instructions - ragcode mcp, ⚖️ the golden rule, project overview, architecture & patterns and developer workflows.
ZipAI CLAUDE.md
Claude Code instructions for nickdesi/ZipAI, covering claude.md — zipai: ultra-dense token optimizer, rules, 1. zero filler, 2. ambiguity and 3. prompt caching.
gpu-ai-skills CLAUDE.md
Claude Code instructions for intel/gpu-ai-skills, covering claude.md, what this repository is, repository structure, commands and validation (required before any skill change).