codex-reference

A reference guide for handing deterministic, repeatable work to a separate executor and collecting compact results. It covers when to delegate, how to launch the work, and how to read the output.

In plain words
What is it for?
Batch parsing, extracting data from disassembly or decompiled code, formatting symbol tables, scanning with YARA rules, and comparing binaries.
Why use it?
It keeps large extraction or processing results from filling the main agent’s working context. It also provides a consistent process for delegated jobs.

Skill for Claude CodeCodex

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 skills/ybx-stack/rich-brain-clean-hands/codex-reference
Any agent
npx skills add ybx-stack/rich-brain-clean-hands --skill codex-reference
Clone the repo
git clone --depth 1 https://github.com/ybx-stack/rich-brain-clean-hands

Made for: Claude Code, Codex.

Per session 0 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,265 The whole file, excluding the scripts and references it only reads on demand.
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 $0.00000 $0.03265
Opus 5 $0.00000 $0.01632
Sonnet 5 $0.00000 $0.00653
Haiku 4.5 $0.00000 $0.00327

Measured 2d ago against content hash 0e9a739e7062, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

codex-reference 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.

examples/codex-reference/SKILL.md · 359 lines

How it starts

The opening of the file, as written. The whole thing — 359 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Clean Hands Delegation — Full Reference Implementation

This is the production-grade SKILL file that the rich brain loads when it needs to delegate. It covers the complete decision tree: when to delegate, who pulls data, how to launch, how to wait, how to read results. Paths use <WORKSPACE> / <TARGET> placeholders — adapt to your layout.

The executor is pluggable. This reference uses a CLI executor, but any process that reads a packet and writes a compact result works. See docs/EXECUTORS.md for alternatives.


When to Use

Delegate when the task meets ALL of these criteria:

  1. Input is deterministic (same input = same output every time)
  2. Does not require interactive judgment, browser, MCP, or real-time decision making
  3. Would produce large output that pollutes rich brain context
  4. Is primarily extraction, formatting, parsing, or batch processing

High-frequency delegation targets

Software reverse engineering:

  • Batch extract function signatures, strings, xrefs from disassembly output
  • Parse large decompilation output by function granularity
  • Symbol table / import-export table extraction and formatting
  • YARA rule batch scanning
  • Binary diff (patch diffing) between versions
  • Large disassembly listing filtering and structuring

Web / JS reverse engineering:

  • Webpack/Vite bundle splitting and endpoint extraction
  • Batch grep API routes, encryption functions, token logic from JS files
  • Source map restoration and batch code analysis
  • HAR/traffic file batch processing and classification
  • Large response body parsing and normalization

General:

  • Log file analysis and pattern extraction
  • Configuration audit across multiple files
  • Dependency tree analysis
  • Large CSV/JSON dataset transformation

DO NOT delegate when

  • Browser interaction needed (breakpoints, hooks, CAPTCHA)
  • MCP tools needed (IDA MCP, browser automation MCP, Burp MCP)
  • Strategy decision needed ("what should we try next?")
  • Output is small enough to handle inline (< 50 lines)

Read the full file on GitHub · 359 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. 2d ago First seen · 359 lines · 0 tokens per session scan A 0e9a739e7062

Subscribe to this mod's changes

codex-reference is a skill published in the GitHub repository ybx-stack/rich-brain-clean-hands (2 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,265 tokens. 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.