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/roninforge/budgetclaw/agents-mdgit clone --depth 1 https://github.com/RoninForge/budgetclawWhat 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 | $0.01627 | $0.01627 |
| Opus 5 | $0.00813 | $0.00813 |
| Sonnet 5 | $0.00325 | $0.00325 |
| Haiku 4.5 | $0.00163 | $0.00163 |
Grade A, and why
budgetclaw 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 — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md - BudgetClaw
Guidance for AI coding agents working on this repository. Human contributors should read CONTRIBUTING.md first; this file exists to give agents the same context without spelunking.
What this is
BudgetClaw is a local spend monitor for Claude Code. It reads the JSONL session logs Claude Code writes under ~/.claude/projects/, attributes each tool call's token cost to a {project, branch} pair, enforces budget caps by sending SIGTERM to the claude process on breach, and pushes phone alerts via ntfy.
Single static Go binary. MIT licensed. Part of the RoninForge toolkit.
Trust pledge (load-bearing - never break this)
The whole product hinges on three claims. Never write code, copy, or comments that contradict them.
- Zero keys. BudgetClaw never reads, stores, or uses an Anthropic API key.
- Zero prompts. BudgetClaw reads only the
usage.*andcwdfields from JSONL log lines. Never the message bodies. - Zero latency added. BudgetClaw is a local log reader. It never sits between an editor and the Anthropic API.
Never call BudgetClaw a "proxy." That word is legally dead under Anthropic ToS §D.4. Always say "usage guard," "spend monitor," or "telemetry reader." If you find existing copy or comments using "proxy," fix them.
Layout
cmd/budgetclaw/ Cobra entrypoint
internal/cli/ Subcommand handlers (init, limit, status, watch, alerts, unlock, pricing, backfill)
internal/db/ SQLite (modernc.org/sqlite, no cgo) - events + rollups
internal/pipeline/ JSONL parser + token-to-cost mapper
internal/pricing/ Embedded pricing table (Opus 4.x, Sonnet, Haiku, cache multipliers)
internal/watcher/ fsnotify-based file tailer
internal/budget/ Limit evaluation + SIGTERM
internal/notify/ ntfy push client
examples/config.toml Documented config template (TOML, not YAML)
testdata/ Real-shape JSONL fixtures
.github/workflows/ CI + daily pricing audit
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 · 103 lines · 1,627 tokens per session scan A eb2d1069ba0d
budgetclaw AGENTS.md is an instructions file published in the GitHub repository RoninForge/budgetclaw (8 stars, last pushed 2d ago), licensed MIT. It adds 1,627 tokens to every session, about $0.0081 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
agentmeter CLAUDE.md
Claude Code instructions for ymstar/agentmeter, covering claude.md, what is agentmeter?, commands, architecture and data flow.
open-agent-hub AGENTS.md
AGENTS.md instructions for guanyang/open-agent-hub, covering agents.md, 1. think before coding, 2. simplicity first, 3. surgical changes and 4. goal-driven execution.
agents
Always-loaded project anchor. Read this first. Contains project identity, non-negotiables, commands, and pointer to ROUTER.md for full context.
gmgn-skills CLAUDE.md
Instructions for GMGNAI/gmgn-skills, covering claude.md, critical rule — read this first, project overview, available skills and quick decision guide.
core copilot-instructions.md
Instructions for asdecided/core: These decisions are already accepted. Do not re-open or contradict them; ask the AsDecided MCP tools (getartifact, searchartifacts) for the full text before proposing a change that touches one.
Atomic-Spec AGENTS.md
Instructions for Chappygo-OS/Atomic-Spec, covering agents.md — adding a new ai agent to atomic spec, about atomic spec and the atomicspec cli, general practices, adding new agent support and current supported agents.