figure-qa
/figure-qa
Use this skill to QA a scientific figure for journal compliance, alignment, palette correctness, and text legibility. Triggers on "QA this figure", "check this figure", "review my figure", "is this figure paper-ready", "validate figure", "check the text in this figure", or when invoked after a figures skill produces output (unless the caller passes no-qa). Dispatches on input type (SVG, raster PNG/JPG/TIFF, Python plot script, or composed-figure directory), runs the programmatic checks including OCR of expected strings and palette compliance against the project theme, adds a vision rubric pass
- Author
- neuromechanist ↗
- Repository
- research-skills ↗
- Licence
- BSD-3-Clause
- Last repository change
- 23.10.2026
Assessment and security
- Files in this package
- 3
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 23.10.2026, without references to retired models or APIs.
BSD-3-Clause, otwarta.
What it does and when to use it
Use this skill to QA a scientific figure for journal compliance, alignment, palette correctness, and text legibility. Triggers on "QA this figure", "check this figure", "review my figure", "is this figure paper-ready", "validate figure", "check the text in this figure", or when invoked after a figures skill produces output (unless the caller passes no-qa). Dispatches on input type (SVG, raster PNG/JPG/TIFF, Python plot script, or composed-figure directory), runs the programmatic checks including OCR of expected strings and palette compliance against the project theme, adds a vision rubric pass
Triggers from the description
Original author description
Use this skill to QA a scientific figure for journal compliance, alignment, palette correctness, and text legibility. Triggers on "QA this figure", "check this figure", "review my figure", "is this figure paper-ready", "validate figure", "check the text in this figure", or when invoked after a figures skill produces output (unless the caller passes no-qa). Dispatches on input type (SVG, raster PNG/JPG/TIFF, Python plot script, or composed-figure directory), runs the programmatic checks including OCR of expected strings and palette compliance against the project theme, adds a vision rubric pass
How it works
- When to use
- What to pass to the reviewer
- Why a fresh-context reviewer
- Dispatch
- The report and the loop
- The brain (do not duplicate into dispatch or agent shells)
SKILL.md preview
---
name: figure-qa
description: Use this skill to QA a scientific figure for journal compliance, alignment, palette correctness, and text legibility. Triggers on "QA this figure", "check this figure", "review my figure", "is this figure paper-ready", "validate figure", "check the text in this figure", or when invoked after a figures skill produces output (unless the caller passes no-qa). Dispatches on input type (SVG, raster PNG/JPG/TIFF, Python plot script, or composed-figure directory), runs the programmatic checks including OCR of expected strings and palette compliance against the project theme, adds a vision rubric pass, and returns a report with a machine-readable JSON verdict that drives the generate, QA, fix loop.
version: 0.2.0
---
# Figure QA
Routes a scientific figure to an independent, fresh-context QA reviewer that detects the input type, runs the deterministic checks (fonts, palette, geometry, alpha, resolution, rendered text) with the helper scripts, and adds a vision-language model (VLM) rubric judgment for the aesthetic dimensions.
This skill is a thin dispatcher: the detection logic, exit-code contract, VLM rubric, report shape, and the iterate loop all live in `references/`, and the deterministic engine lives in the figures plugin's `agents/figure-qa-scripts/`.
## When to use
Activate when a figure needs a journal-submission QA pass, or after a figures skill (`figures:scientific-figure`, `figures:svg-figure`, `figures:svg-primitives`, `figures:transparent-icons`, `figures:ai-full-figure`, `figures:plot-styling`) produces output, unless the caller passes `no-qa`.
## What to pass to the reviewer
Every dispatch carries the same briefing, so the right checks run:
- the figure path and, when known, the input type
- the target journal (`nature`, `science`, `cell`, `pnas`, `poster`, `slide`, or `generic`)
- the theme path (`figures/theme.json`) so palette compliance is judged against the project bible rather than a fixed allow-list
- every verbatim string the figure was asked to render (titles, panel letters, labels); without this list the raster text check is silently off
- the physical width in millimetres for raster inputs, so text height converts to points
## Why a fresh-context reviewer
A QA pass is more trustworthy from a reviewer that did not just author the figure.
Run it in a separate context and pass only the briefing above.
On tools without subagents, run the procedure inline.
For several candidates, issue one dispatch per candidate in a single message so they run in parallel on the Sonnet tier.
## Dispatch
In every branch the reviewer follows `references/figure-qa-procedure.md` (strict separation: scripts own ground-truth measurements, the VLM owns aesthetic judgment).
- **Claude Code:** `Agent(subagent_type: "figures:figure-qa", ...)` with the briefing above. Honor a `no-qa` opt-out by returning immediately.
- **Codex CLI:** plugin installation exposes this skill, not a Codex subagent. To use a fresh-context Codex reviewer, first copy `${CLAUDE_PLUGIN_ROOT}/agents/templates/figure-qa.toml` (the plugin's `agents/templates/` directory) to `~/.codex/agents/` or `.codex/agents/`, then invoke that configured agent if the current Codex surface supports `/agent`. If no Codex subagent is configured or available, use the fallback branch.
- **Copilot CLI:** plugin installation exposes this skill and, through `.github/plugin/plugin.json`, the `.agent.md` reviewer in `agents/templates/`. Invoke that configured agent when the current Copilot surface supports custom agents. If running outside a plugin install, copy `agents/templates/figure-qa.agent.md` to `.github/agents/` or `~/.copilot/agents/`. If no custom agent is available, use the fallback branch.
- **Fallback** (no subagent support, or an interactive in-thread check): first locate the procedure (`$CLAUDE_PLUGIN_ROOT/skills/figure-qa/references`, else `find . -type d -path '*/skills/figure-qa/references' | head -1`); if it cannot be found, stop and tell the user to install the figures plugin rather than guessing checks. Then follow `references/figure-qa-procedure.md` directly.
## The report and the loop
The report keeps its markdown sections and ends with one fenced JSON block: `status` (`ship`, `revise`, `block`), `findings[]` each with `check`, `severity`, `message`, `action` (`regenerate`, `edit`, `overlay`, `rescale`, `recolor`, `none`) and a `hint`, plus `measurements` and the five VLM scores.
Generation skills branch on that block.
`references/iterate-loop.md` defines the generate, QA, fix, regenerate loop (N candidates in parallel, one targeted change per iteration, stop at `ship` or after three iterations) that `figures:ai-full-figure` follows.
## The brain (do not duplicate into dispatch or agent shells)
- `references/figure-qa-procedure.md`: no-qa opt-out, script location, type detection, per-branch checks including the raster text check and theme palette compliance, exit-code contract, VLM rubric, the exact report shape, and the finding-to-action table.
- `references/iterate-loop.md`: the candidate, QA, fix loop with its worker briefing and stopping conditions.
- `agents/figure-qa-scripts/check_{svg,raster,plot_script}.py` (in the figures plugin): the deterministic engine, kept in place so other skills can call it directly. `check_raster.py --json --expect-text "..." --palette figures/theme.json --width-mm 89 --journal nature` is the text and palette check; `check_svg.py --json --palette figures/theme.json` covers SVG geometry, fonts, and palette.
Files in this package (3)
- SKILL.md 5 KB
- references/figure-qa-procedure.md 16 KB
- references/iterate-loop.md 5 KB
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · neuromechanist/research-skills | 90 | BSD-3-Clause |
| raoofaltaher/science-ultrapowers | 90 | MIT |
Database snapshot: 13.11.2026 · package version f0ce94f0fed8 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100figure-bible
Use this skill when the user asks to "make a figure bible", "set up a style bible", "create a figure theme", "make a theme for my figures", "set up a figure style", "palette for my paper figures", or wants a consistent figure style. Scaffolds and validates figures/theme.json;… - 100scientific-figures
Create, style, export, and validate publication-quality scientific and technical figures. xy is the preferred default (matplotlib.pyplot-compatible for easy migration); matplotlib/seaborn/plotly also supported. Use when the user asks for a "publication figure", "paper figure",… - 100how-to-review-figure
Agentic trigger. Use when judging the visual design of a figure, plot, or diagram against a scientific-plot rubric. - 90grant-figure-qa
Use this skill to review grant proposal figures for compliance, resolution, accessibility, and quality. Triggers on "check grant figures", "review proposal figures", "figure QA for grant", "NIH figure requirements", or when preparing a grant for submission. - 100figure-composer
Compose or improve a publication-grade multi-panel scientific figure from a claim, concrete data paths, or an existing image. Use for figure outlining, parallel panel rendering, exact-grid composition, visual inspection, and adversarial figure review. Use figure-style for one… - 100tavotto-figure
画 matplotlib 论文级图表,并用 Tavotto 继续微调(拖图例、改字号线宽、调刻度、按出版规范预检、导出矢量 PDF);也把用户**已经画好的图**按投稿要求改宽度 / 字体 / 最小字号而不重画(保留式规范化)。用户要画图、出图、做 figure、画折线/柱状/散点/误差棒、做论文配图、scientific plot / publication figure,要把现有图改成 8 cm / 换 Times New Roman / 字号不小于 8 pt,或提到 Tavotto 时使用。