sonicjs AGENTS.md

A repository guide for SonicJS, including rules for exploring code, delegating investigations, and using project-specific tools. It describes codegraph, an index that helps locate symbols and trace calls across the codebase, and rtk, a tool that shortens shell output.

In plain words
What is it for?
Use it to locate symbols, trace how code flows, find callers and called functions, assess the impact of a change, or investigate broad cross-cutting questions.
Why use it?
It gives coding agents a consistent way to find code and understand dependencies in a large repository.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/sonicjs-org/sonicjs/agents-md
Clone the repo
git clone --depth 1 https://github.com/SonicJs-Org/sonicjs

Made for: Codex, OpenCode.

Per session 2,447 This file is loaded in full into every session.
When invoked 2,447 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.02447 $0.02447
Opus 5 $0.01223 $0.01223
Sonnet 5 $0.00489 $0.00489
Haiku 4.5 $0.00245 $0.00245

Measured 3d ago against content hash f5cd8b71b8fb, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

sonicjs 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 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.

AGENTS.md · 139 lines

How it starts

The opening of the file, as written. The whole thing — 139 lines — stays where its author put it; the contents beside it link to each section on GitHub.

SonicJS Agent Guidelines

These instructions mirror the workflow Claude Code already follows in Conductor so every coding agent behaves consistently.

Token-Efficient Tooling (use FIRST, before grep/Read sprees)

This repo is indexed by codegraph (684 files, 5693 nodes). Use it as the default code-lookup tool. Hook also rewrites shell commands via rtk for 60-90% bash savings. Default response mode is caveman for terse output.

Decision tree — code exploration

Intent Tool Why
"How does X work / where does X live / trace flow X→Y" mcp__codegraph__codegraph_explore ONE call returns verbatim source grouped by file. Replaces many Read/Grep.
"Just the location of symbol named X" mcp__codegraph__codegraph_search Cheapest — name + path only.
"What calls this / what does this call / blast radius" codegraph_callers / codegraph_callees / codegraph_impact Edges already indexed; grep can't follow dynamic dispatch.
Pinpoint known file edit Read + Edit Codegraph not needed for known path.
Truly open-ended cross-cutting search Agent w/ Explore or cavecrew-investigator Compressed output, protects main context.

Rule: Before any 3+ Grep/Read combo for "where is X", call codegraph_explore first. Treat codegraph as a pre-built index — don't re-run searches it already answers.

rtk (Rust Token Killer)

  • Bash commands auto-rewritten by hook — transparent, 0 overhead. No action needed.
  • rtk gain to audit savings; rtk discover to find missed opportunities.
  • Don't bypass rtk with raw shell unless debugging.

caveman mode

  • Active by default (full). Drops articles/filler. Code blocks, commits, security warnings stay normal.
  • Toggle: /caveman lite|full|ultra, off via "stop caveman".

Delegation — cavecrew subagents

Use when scope is bounded AND output would bloat main context:

  • cavecrew-investigator — read-only locator ("where is X / what calls Y / map this dir"). ~60% smaller tool result vs vanilla Explore.
  • cavecrew-builder — surgical 1-2 file edit (typo, single-fn rewrite, mechanical rename). Refuses 3+ files.
  • cavecrew-reviewer — diff/PR review, one line per finding.

Read the full file on GitHub · 139 lines

Changes

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.

  1. 3d ago First seen · 139 lines · 2,447 tokens per session scan A f5cd8b71b8fb

Subscribe to this mod's changes

sonicjs AGENTS.md is an instructions file published in the GitHub repository SonicJs-Org/sonicjs (1,693 stars, last pushed 5d ago), licensed MIT. It adds 2,447 tokens to every session, about $0.0122 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.