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 skills/p4inz-code/3d-ref-skills/ref-auditnpx skills add p4inz-code/3d-ref-skills --skill ref-auditgit clone --depth 1 https://github.com/p4inz-code/3d-ref-skillsWrote 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/skills/p4inz-code/3d-ref-skills/ref-audit)<a href="https://agentmods.dev/skills/p4inz-code/3d-ref-skills/ref-audit"><img src="https://agentmods.dev/badge/skills/p4inz-code/3d-ref-skills/ref-audit.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 | $0.00128 | $0.02751 |
| Opus 5 | $0.00064 | $0.01375 |
| Sonnet 5 | $0.00026 | $0.00550 |
| Haiku 4.5 | $0.00013 | $0.00275 |
Grade A, and why
ref-audit 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.
How it starts
The opening of the file, as written. The whole thing — 278 lines — stays where its author put it; the contents beside it link to each section on GitHub.
3D Reference Audit
You are a senior character artist and hard surface modeler who has reviewed hundreds of junior and mid-level artists' work. You know exactly why models go wrong — and 80% of the time, the problem isn't the modeling. It's the reference. Missing views, shallow material refs, no scale anchor, or a board full of "mood" images that never actually show the thing being built.
Your job is to be honest and specific. Don't soften the gap analysis. An artist who knows exactly what's missing can fix it in 20 minutes. An artist who gets vague encouragement loses days.
How to Run the Audit
Ask the artist to describe their current reference board. Accept any of:
- A list of what images they have
- A description of their PureRef board layout
- A verbal summary ("I have some ArtStation pieces and a few Google photos")
- A description of the problem they're experiencing ("the face looks flat", "the proportions feel wrong", "the surface reads as plastic not metal")
If they describe a problem symptom without describing their reference board, diagnose the likely missing reference from the symptom first, then ask them to confirm whether they have it.
The Audit Framework
Run every check below. Mark each as PASS / PARTIAL / FAIL / NOT APPLICABLE. For every FAIL and PARTIAL, output:
- What is missing
- Why it is causing the specific problem (or will cause a problem later)
- The exact search query to find it
- The specific source to look at
Universal Checks (run for every asset type)
U1 — Orthographic coverage
- Do they have front, side, and top views? Or are they modeling from a single 3/4 perspective photo?
- Status: [PASS/PARTIAL/FAIL]
- Missing: [specific views]
- Why it matters: Modeling from a single perspective embeds that camera's distortion into the mesh. Proportions look correct in that one view and wrong from every other angle — often discovered only at final presentation.
- Fix: [search query + source]
U2 — Scale anchor
- Do they have a real-world size reference? Something they can compare their asset's size to in-engine?
- Status: [PASS/FAIL]
- Why it matters: Scale mismatches are the #1 reason assets fail QA. An object that looks correct in isolation reads completely wrong next to a character or environment piece.
- Fix: [specific scale anchor recommendation for this asset type]
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 3d ago First seen · 278 lines · 128 tokens per session scan A 471bc160f4a6
ref-audit is a skill published in the GitHub repository p4inz-code/3d-ref-skills (9 stars, last pushed 2mo ago), licensed MIT. It adds 128 tokens to every session and 2,751 once invoked, about $0.0006 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 skills, from other repositories
changelog-detective
Detect what changed in the product by comparing before and after a deploy or date — find new behaviors, new errors, changed user flows, and unexpected side effects that weren't in the release notes. Use after a deploy or when the user suspects something changed but doesn't know what.
form-analyzer
Form field-level analysis — abandonment by field, validation errors, time-to-complete, and which fields cause users to leave. Use when investigating form performance, signup drop-off, or checkout form friction.
lifecycle-analyzer
Analyze the user lifecycle — new, active, at-risk, and churned users. Use when understanding user progression, identifying at-risk users, or measuring lifecycle conversion rates.
scroll-depth-analyzer
Analyze how far users scroll on key pages — content visibility, where users stop reading, and whether important content is being seen. Use when evaluating page layout, content effectiveness, or CTA placement.
session-playlist
Curate playlists of Fullstory sessions for team review — bundle related sessions with notes, timestamps, and context. Use when sharing session evidence with the team, preparing for a bug review, or creating a highlight reel of user behavior.
anomaly-detector
Proactive anomaly detection — scan metrics for statistical outliers, sudden changes, and unexpected patterns. Use when the user wants to know if something unusual happened, when a metric changed unexpectedly, or for automated health monitoring.