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/arjia-labs/sark/agents-mdgit clone --depth 1 https://github.com/Arjia-Labs/sarkWrote 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/arjia-labs/sark/agents-md)<a href="https://agentmods.dev/instructions/arjia-labs/sark/agents-md"><img src="https://agentmods.dev/badge/instructions/arjia-labs/sark/agents-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 | $0.02580 | $0.02580 |
| Opus 5 | $0.01290 | $0.01290 |
| Sonnet 5 | $0.00516 | $0.00516 |
| Haiku 4.5 | $0.00258 | $0.00258 |
Grade A, and why
sark AGENTS.md scanned grade A with 1 finding 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 4d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s https://sark.<subdomain>.workers.dev/health How it starts
The opening of the file, as written. The whole thing — 249 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — deploying sark
A runbook for an AI agent asked to clone, configure, and deploy this Cloudflare Worker. Follow the phases in order. Each ends with a check; do not continue past a failed check.
Some steps need a decision only the human can make. Those are marked ASK — stop, ask, and wait. Do not guess a Slack channel id, a billing plan, or an allowlist.
Rules
- Never print a secret value. Not in output, not in a commit, not in a comment. Pipe secrets through stdin. If one is ever echoed, say so and tell the human to rotate it.
- Never commit
.dev.varsor.env. Both are gitignored. Keep it that way. - Never put personal settings in
wrangler.jsonc. It is the public template. Worker name, origin, allowlist, and template box id go in.deploy.env(gitignored) and are injected bynpm run deploy.account_idcomes fromCLOUDFLARE_ACCOUNT_ID.npm run check-configenforces this, and it runs in CI. - If a command fails twice the same way, stop and report. Do not improvise around the Cloudflare or Box APIs.
Phase 0 — prerequisites
node --version # must be >= 22
git --version
npx wrangler --version
ASK the human to confirm all three before you touch anything:
- A Cloudflare account on a paid Workers plan. The per-thread state machine is a Durable Object with a SQLite backend, which the free plan does not offer. Deploy will fail at the migration step otherwise.
- An ascii.dev Box account and a
box_...API key. This Worker is a control plane for Box sandboxes and does nothing without one. - Whether Slack is wanted at all. It is optional; the
/apisurface exercises the entire pipeline without a Slack app. If they say no, skip Phase 5.
Authenticate to Cloudflare, whichever the human prefers:
npx wrangler login # interactive, opens a browser
# or export CLOUDFLARE_API_TOKEN=... # a token with Workers Scripts:Edit
export CLOUDFLARE_ACCOUNT_ID=... # required either way
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.
- 4d ago First seen · 249 lines · 2,580 tokens per session scan A fc78359702ca
sark AGENTS.md is an instructions file published in the GitHub repository Arjia-Labs/sark (1 stars, last pushed 1mo ago), licensed MIT. It adds 2,580 tokens to every session, about $0.0129 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
vibesdk AGENTS.md
AGENTS.md instructions for cloudflare/vibesdk, covering agents.md, tooling, verification, frontend ui and frontend data fetching.
vibesdk CLAUDE.md
Claude Code instructions for cloudflare/vibesdk, covering claude.md, communication style, project overview, key architectural patterns and common development tasks.
cloudbox AGENTS.md
Instructions for acoyfellow/cloudbox, covering cloudbox agent guide, canonical commands, deploy model, cloudflare account boundaries and required production secrets.
kubefwd CLAUDE.md
Claude Code instructions for txn2/kubefwd, covering claude.md, project overview, integration and usage patterns, tui (terminal user interface) and tui features.
saasmail CLAUDE.md
Claude Code instructions for choyiny/saasmail, covering saasmail, development and skills.
baerly-storage CLAUDE.md
Toolchain, verification matrix, module map, conventions, anti-patterns. The main agent entry point.