Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →jpvarbed/visual-critique
Run three independent Codex vision inspections of a render, pose, screenshot, or UI, then synthesize a consensus report. Use to verify visible anatomy, geometry, alignment, or polish before proceeding; use design-critique for structured UX feedback and adversarial-review for text artifacts.
Passed 18 of 21 test cases with the skill, and 10 of 21 without it.
and 6 more in the full report
9 fixed11 passed either way3 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
When attempting to run `skills/review/visual-critique/visual-critique.sh` with `--focus "check shadows"`, the script fails with the error message `cannot read image path`. What missing argument causes this error?
| Checked | Without skill | With skill |
|---|---|---|
| explains that an image path must be passed after the focus parameter | Missed | Passed |
The instructions your model reads when the skill loads.
Run three read-only vision inspections against the same image, then use a fourth Codex pass to keep findings supported by at least two inspections. Gemini is prohibited from this workflow.
skills/review/visual-critique/visual-critique.sh \
--focus "check pelvis tilt, knee flex, and spine continuity" \
path/to/render.png
The script attaches the image to three parallel codex exec calls in a read-only sandbox. If any
inspection fails or returns empty output, the review fails instead of synthesizing incomplete
evidence. The final pass receives only the three textual inspections and returns the consensus.
| Issue | Fix |
|---|---|
codex CLI not found on PATH | Install/authenticate Codex and confirm codex --version. |
cannot read image path | Pass one readable image path after --focus. |
| An inspection returns nonzero or empty | Read the emitted engine error, fix Codex/auth/capacity, and rerun all three inspections. |
| The model cannot decode the image | Convert it to PNG first, e.g. sips -s format png in.heic --out in.png. |
test cases passed, vs 10 / 21 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.