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/ntorga/agent-starter-kit/agents-mdgit clone --depth 1 https://github.com/ntorga/agent-starter-kitWhat 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 | $0.03255 | $0.03255 |
| Opus 5 | $0.01628 | $0.01628 |
| Sonnet 5 | $0.00651 | $0.00651 |
| Haiku 4.5 | $0.00326 | $0.00326 |
Grade A, and why
agent-starter-kit AGENTS.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 2d 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 — 147 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
A style book for code that reads itself. Internalize the reasoning, apply the judgment, write code that the next reader can follow without reverse-engineering.
Honesty and Ambiguity
If a request is flawed, say so — agreeable silence produces bad code. When the user assigns a task, proceed on any non-destructive reversible action needed to complete it; git is the safety net, stopping to confirm the obvious wastes the session. Stop only for destructive or irreversible actions (delete repository, drop database, force-push to main), or genuinely non-obvious trade-offs. When intent could mean different things, escalate with structure: one problem, three options with trade-offs, one recommendation with reason. When the ambiguity only changes how (not what), proceed inline with a documented default.
Code Shape
Getting function boundaries right means the reader never reassembles scattered logic or untangles mixed concerns. Three questions: what to extract, what to unify, what to keep together.
Function Extraction
Extract a function when it performs a transformation through genuine logic — input processed into distinct output. Inline a function when its body delegates to a single call with trivial wrapping (rename, forward, reformat). A 60-line function performing one operation is correct; a 5-line function that wraps another call is indirection. The test: removing the function and inlining its body at the call site — does the caller lose clarity or gain it?
Duplication
Two code paths that share knowledge require synchronized edits — a change to one demands the same change to the other. Unify them. Two code paths that look similar but answer different questions are two responsibilities that happen to resemble each other. Shared code couples them; when one question evolves the other drags. The test: would a requirement change affect both paths identically? Identical → unify. Divergent → the resemblance is coincidental, keep separate.
Single Responsibility
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.
- 2d ago First seen · 147 lines · 3,255 tokens per session scan A 4a7a8197e35e
agent-starter-kit AGENTS.md is an instructions file published in the GitHub repository ntorga/agent-starter-kit (140 stars, last pushed 1mo ago), licensed MIT. It adds 3,255 tokens to every session, about $0.0163 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
open-multi-agent AGENTS.md
AGENTS.md instructions for open-multi-agent/open-multi-agent, covering agents.md, repository map, commands, working rules and validation by change type.
agent-framework copilot-instructions.md
Instructions for microsoft/agent-framework, covering github copilot instructions, repository structure and architectural decision records (adrs).
agent-framework python.instructions.md
Instructions for microsoft/agent-framework: See AGENTS.md for project structure and package documentation. Detailed conventions are in the agent skills under .github/skills/.
open-multi-agent CLAUDE.md
Claude Code instructions for open-multi-agent/open-multi-agent, a project described as: TypeScript AI agent orchestration framework with dynamic workflows. Describe the goal, not the graph: a coordinator plans the task DAG at runtime and runs it on any LLM (Claude, ChatGPT, Gemini, DeepSeek, or local models).
kdcube AGENTS.md
Instructions for kdcube/kdcube, covering agents.md — operating rules for coding agents, shared ground rules (both classes), a. platform contributors, git and shared-tree etiquette and code architecture.
squads-cli AGENTS.md
Instructions for agents-squads/squads-cli, covering agent instructions, specs — read before non-trivial changes, what this is, repository structure and build and test.