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/victortomaili/agent-cli/agents-mdgit clone --depth 1 https://github.com/VictorTomaili/agent-cliWhat 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 | $0.00950 | $0.00950 |
| Opus 5 | $0.00475 | $0.00475 |
| Sonnet 5 | $0.00190 | $0.00190 |
| Haiku 4.5 | $0.00095 | $0.00095 |
Grade A, and why
agent-cli 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 — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Instructions for any AI coding agent (Claude Code, Codex, Cursor, etc.) working in this repo. Humans: see README.md for the user-facing docs and ARCHITECTURE.md for the full layer-by-layer map.
What this is
agent-cli (npm: @victortomaili/agent-cli, bin: agent-cli) manages AGENTS.md-style
instruction files and syncs them across every AI coding tool a user has installed. Node.js
= 22, ESM (
"type": "module"), zero build step —src/cli.jsruns directly.
Install / update / init
These are the commands a user (or an agent driving npm/agent-cli on a user's behalf) runs
against the published package, not this checkout:
npm install -g @victortomaili/agent-cli # install
npm update -g @victortomaili/agent-cli # update to latest published version
agent-cli init # bootstrap ~/AGENTS.md + pointer stubs (idempotent)
Don't confuse that npm update with agent-cli update / agent-cli upgrade — the latter apply
shipped-default content (seed files, managed instruction blocks) to an already-installed
user's ~/.agents/ brain and are unrelated to the package version.
Working in this checkout
npm install # dependencies
npm link # global `agent-cli` command backed by this checkout (for manual testing)
npm run check # syntax-check every src/*.js file (= npm run lint = npm run build)
npm test # node --test test/*.test.js
Run npm run check && npm test before considering any change done — CI (.github/workflows/ci.yml)
runs the same two commands on the ubuntu/windows × Node 22/24 matrix, and .github/workflows/publish.yml
publishes to npm automatically on every main push whose package.json version isn't on the
registry yet, so a broken check/test blocks a real release, not just a PR.
Conventions to preserve
- Layering (
src/cli.js→src/commands/*.js→src/*.jslib;src/api/index.jsis a read-only SDK over lib;src/skills/**is self-contained). Enforced bytest/import-boundaries.test.js— read it before adding a new cross-module import. - DI pattern: every command module exports
registerXCommands(program, deps);src/cli.jsbuildsdepsonce. Keep command files thin — parse options, call lib functions, format output. - Cross-cutting invariants (see ARCHITECTURE.md's "do not weaken these"): atomic writes
(
util.js), locked config writes (cross-process CAS inconfig.js), path containment on any write/delete driven by untrusted input, pointer-only deletion inunlink, secrets never synced/snapshotted/searched. - npm publish surface: only what's listed in
package.json's"files"array (src,seed,README.md,LICENSE,docs/contract.md) ships in the tarball. Note the allowlist namesdocs/contract.mdas a single path, not thedocs/folder: everything else underdocs/is a local working analysis (gitignored, see.gitignore) and must never ship.ARCHITECTURE.mdand.spec/never do — don't assume a new root-level doc is visible to an installed user. .spec/— single local-only folder for the project's spec-driven-dev workflow AND internal planning notes (README.md,constitution.md,specs/,plans/,tasks/,templates/,ROADMAP.md,findings.md). Gitignored, never committed, never shipped in the tarball. The CLI'sspectcommand points here. Don't recreate these docs at the repo root.
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 · 68 lines · 950 tokens per session scan A bfe9987f6b98
agent-cli AGENTS.md is an instructions file published in the GitHub repository VictorTomaili/agent-cli (1 stars, last pushed 6d ago), licensed MIT. It adds 950 tokens to every session, about $0.0047 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 2026-08-31.
Other instructions, from other repositories
Tool.Agent-Harness-Deploy AGENTS.md
Instructions for masteryee-labs/Tool.Agent-Harness-Deploy, covering agent harness deploy — agent entry file, 0. what this repo is (read this before doing anything), 1. boot protocol (order is mandatory), 2. the deploy contract (what "幫我部屬" means) and 3. red lines (violating any → stop and ask the human).
vibecode-pro-max-kit CLAUDE.md
Instructions for withkynam/vibecode-pro-max-kit, covering claude.md, bootstrap guard, before any substantial task, riper-5 spec-driven development system and shared development protocols.
unit-tests-skills CLAUDE.md
Instructions for mavka-ai/unit-tests-skills, covering claude.md, what this repo is, repository structure, available skills and workflow.
vibecode-pro-max-kit AGENTS.md
Instructions for withkynam/vibecode-pro-max-kit, covering agents.md, riper-5 spec-driven development system, shared development protocols, orchestrator role (main codex session) and repository context.
unit-tests-skills AGENTS.md
Instructions for mavka-ai/unit-tests-skills, covering agents.md, skills, rules location, plugin validation and what it checks, and why each check exists.
ai-agent-instruction-modules AGENTS.md
Instructions for tablackburn/ai-agent-instruction-modules, covering ai agent instructions, instructions for ai agents, instruction applicability matrix, available instruction files and quick reference.