sm64ds-decomp: Skill for Claude Code

.claude/skills/decomp-match-review/SKILL.md

decomp-match-review is a skill for Claude Code from tangosdev/sm64ds-decomp. It costs 146 tokens per session (8,394 once invoked), scanned A, original, MIT.

A review guide for changes in a video-game decompilation project, where recovered source code is checked against the original game's compiled program. It checks both whether the binary match is genuine and whether the code remains accurate and maintainable.

In plain words
What is it for?
Reviewing or preparing pull requests, branches, and diffs in decompilation projects. It helps assess the match, the recovered source code, and the cost to port it to other systems.
Why use it?
A matching binary can still hide fake matches, incorrect compiler assumptions, or poorly recovered C++ code. This helps reviewers catch those problems before they become part of the project.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: mentions CLAUDE.md; mentions AGENTS.md.

This is tangosdev/sm64ds-decomp's own configuration. It tells Claude Code how to work on sm64ds-decomp itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything sm64ds-decomp configures →

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is python tools/port_refcheck.py.

Reuse

Borrowing it

Nothing to install: this file belongs to tangosdev/sm64ds-decomp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/tangosdev/sm64ds-decomp/main/.claude/skills/decomp-match-review/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/tangosdev/sm64ds-decomp

Made for: Claude Code.

Wrote 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.

agentmods badge for decomp-match-review

README.md
[![agentmods](https://agentmods.dev/badge/skills/tangosdev/sm64ds-decomp/decomp-match-review.svg)](https://agentmods.dev/skills/tangosdev/sm64ds-decomp/decomp-match-review)
Your own site
<a href="https://agentmods.dev/skills/tangosdev/sm64ds-decomp/decomp-match-review"><img src="https://agentmods.dev/badge/skills/tangosdev/sm64ds-decomp/decomp-match-review.svg" alt="Measured on agentmods" height="20"></a>
Per session 146 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 8,394 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
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 5.1 $0.00146 $0.08394
Opus 5 $0.00073 $0.04197
Sonnet 5 $0.00029 $0.01679
Haiku 4.5 $0.00015 $0.00839

Measured 7d ago against content hash 9dfd15c4fc76, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

decomp-match-review 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 7d 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.

.claude/skills/decomp-match-review/SKILL.md · 522 lines

How it starts

The opening of the file, as written. The whole thing — 522 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Decompilation match review

What the project is for — read this before you rank anything

sm64ds-decomp's goal, in the maintainer's words and in this order:

  1. A historically accurate representation of the C++ version of SM64DS, byte-matched to the .nds. One goal, two halves, both required.
  2. Portability.
  3. Readability.

The two halves of goal 1 are not equally defended. The byte match is falsifiable and gated; the C++ accuracy is neither. match.py fails loudly when the bytes are wrong. Nothing fails when a file byte-matches while spelling _ZN9ActorBasenwEj as a C identifier, declaring a destructor int(char*), or leaving a member filed as a free function. That asymmetry is why the C++ half rots, and it is what a human reviewer is for — everything the gates catch, they already caught before you opened the diff.

The pass order below is review sequencing, not this ranking. Pass 0 runs first because it is the cheapest disqualifier and a fake match is a lie in the progress bar — not because bytes outrank structure. Do not read pass order as priority order.

Two consequences worth stating, because they resolve most ranking questions on their own:

  • A byte-free fix that recovers original structure outranks a byte-free readability fix. Both are free; only one delivers goal 1. Renaming sp0index is free and welcome. Correcting extern int _ZN5ActorD2Ev(char*) to the signature the mangled name already states is also usually free — "a pointer and an int are both one register" — and it is primary work. Ask for the second at least as hard as the first. (Usually free, not always: signedness changes instruction selection and a float changes codegen wholesale, so it stays a verify-cycle question — see B1 — never a patch you hand over.)
  • When accuracy and readability conflict, accuracy wins, and this is the common case. Unnatural store orders, odd-looking casts, and static on a callback are frequently what the original contained. The "Do NOT recommend these" section near the end is that principle in list form; it is not a list of exceptions, it is goal 1 outranking goal 3.

Read the full file on GitHub · 522 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. 7d ago First seen · 522 lines · 146 tokens per session scan A 9dfd15c4fc76

Subscribe to this mod's changes

decomp-match-review is a skill published in the GitHub repository tangosdev/sm64ds-decomp (147 stars, last pushed today), licensed MIT. It adds 146 tokens to every session and 8,394 once invoked, about $0.0007 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.

Related

Other skills, from other repositories

godot-code-review

Use when reviewing GDScript or C# Godot code — checklist of best practices, common anti-patterns, and Godot-specific pitfalls.

jame581/GodotPrompter · 33 tokens

animation-review

Reviews Godot Animation implementation for known pitfalls. Triggers AFTER implementation, when code involves AnimationPlayer, AnimationTree, AnimatedSprite2D, SpriteFrames, AnimationNodeStateMachine, BlendSpace, OneShot, callbackmodeprocess, or animation playback control (play/travel/start). Do NOT use this skill for…

valkor-ai/loom · 75 tokens

tilemap-review

Reviews Godot TileMap implementation for known pitfalls. Triggers AFTER implementation, when code involves TileMapLayer, TileSet, TileSetAtlasSource, TileSetScenesCollectionSource, terrain painting, setcell, erasecell, getcelltiledata, collision polygons on tiles, or NavigationRegion2D with tile-based navigation. Do…

valkor-ai/loom · 85 tokens

second-pass-review

Independent audit of sanitized specs in workspace/output/. Three parallel LLM-based reviewer roles check structural leakage, content contamination, and behavioral completeness. Run AFTER Layer 5 sanitization, BEFORE implementation handoff.

prime-radiant-inc/greenfield · 44 tokens

hz-unity-code-review

Reviews Unity code targeting Meta Quest and Horizon OS for performance issues, rendering best practices, and common VR pitfalls. Use during code review or when diagnosing Quest performance problems in Unity projects.

meta-quest/agentic-tools · 42 tokens

pwn-ai-agent-sast

Drive PWN::AI::Agent::SAST from pwneval.

0dayInc/pwn · 22 tokens