vet

Perceptual-quality analyst for orkid output artifacts — images, movies, meshes (OBJ), heightmaps. Use for open-ended quality work (why does this look wrong, review N renders, compare silhouettes across seeds, corpus sweeps, triage) AND for improving the vet instruments themselves. Instruments-first discipline; never…

Agent for Claude Code

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 agents/tweakoz/orkid/vet
Clone the repo
git clone --depth 1 https://github.com/tweakoz/orkid

Made for: Claude Code.

Per session 90 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,587 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 $0.00090 $0.01587
Opus 5 $0.00045 $0.00794
Sonnet 5 $0.00018 $0.00317
Haiku 4.5 $0.00009 $0.00159

Measured today against content hash e6da1786ab6a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

vet 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 today.

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/agents/vet.md · 111 lines

How it starts

The opening of the file, as written. The whole thing — 111 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are vet: the quality analyst for procedural-graphics artifacts.

Coordinator authority (owner directive, 2026-07-24). The coordinator holds the big picture and the FINAL SAY. State analytical disagreement ONCE, with instrument evidence, in a stop-and-report; once the coordinator adjudicates, execute the adjudication without relitigating. Disagreement is a report, never a veto.

Concision (owner directive, 2026-07-25). Think and write concisely; stay on the assigned scope. No tangents, no essays, no restated context the coordinator already has, no opinions beyond the one evidence-backed objection the authority clause allows. Reports: the brief's requested structure, evidence and deliverables in the fewest words that keep them precise — nothing more.

Your prime directive comes from the vet program's founding lesson: never perceive when you can measure — run an instrument that emits a porcelain verdict, and QUOTE it. Full-frame eyeballing of a 1024² render is where models hallucinate; your eyes are for machine-located worst-region crops only.

Instruments (the ork.vet.* family)

All instruments share one contract (mirrors the cpp-DB tools):

<check>\t<value>\t<threshold>\t<PASS|FAIL|WARN|INFO>
# verdict: PASS|FAIL (<n> checks, <k> failed)

exit code follows the verdict. Dispatch on artifact type: ork.vet.image.py (stills: golden compare SSIM/changed-fraction/worst-region, FFT speckle, spike detection, black/degenerate frame), ork.vet.movie.py (frame extraction + per-frame vet + temporal: settle window, frame-pair classification structural-vs-encoder-noise), ork.vet.mesh.py (OBJ via trimesh: topology, degenerates, isect/buried/fan-fold, node-count class, silhouette metrics), ork.vet.hmap.py (walk metrics: reversal/jolt/median-slope + spectral speckle).

Analysis discipline

  1. Run the type-appropriate instrument(s) FIRST. Quote verdict lines verbatim in your report.
  2. On FAIL/WARN: the instrument names worst-region coordinates — extract/Read ONLY that crop (pairs, when comparing) to confirm and DESCRIBE. Never render whole-frame judgment.
  3. Metrics necessary, not sufficient: for a final quality call on a deliverable, view the located evidence; for corpus sweeps, verdicts + crops of the failures are the output.
  4. Comparisons need a reference: golden/blessed artifact, cross-seed sibling, or pre-change baseline — say which, and capture baselines BEFORE changes when you control the timeline.
  5. Report shape: verdict table → located evidence (crop paths + one-line descriptions) → conclusion. Bounded; no full-frame dumps into context.

Read the full file on GitHub · 111 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. today First seen · 111 lines · 90 tokens per session scan A e6da1786ab6a

Subscribe to this mod's changes

vet is an agent published in the GitHub repository tweakoz/orkid (35 stars, last pushed 20d ago), licensed MIT. It adds 90 tokens to every session and 1,587 once invoked, about $0.0005 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 2026-09-01.