mcpbeat

Survey Visuals

willoscar/survey-visuals

| Draft non-prose visuals artifacts (timeline, figure specs) for a survey, grounded in evidence and using citation keys from `citations/ref.bib`.

6k tokens
context cost
the whole folder, loaded on every use
6
files
ships runnable scripts
0
copies elsewhere
how many repositories repackaged it
496
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/WILLOSCAR/research-units-pipeline-skills --skill survey-visuals

What comes with it

19 890 bytes besides the instruction
assets/figure_templates.yaml
references/figure_archetypes.md
references/overview.md
references/timeline_patterns.md
scripts/run.py

The instruction itself

16 sections, as written by the author

Survey Visuals (timeline + figure specs; NO PROSE)

This skill creates non-prose artifacts that make the writing stage less template-y:

  • timeline / evolution bullets
  • figure specs (what to draw, why it matters, what papers support it)

Load Order

Always read:

  • references/overview.md
  • references/figure_archetypes.md

Read by task:

  • references/timeline_patterns.md when building timeline milestones

Machine-readable assets:

  • assets/figure_templates.yaml — figure archetype specifications (extensible without code changes)

Script Boundary

Use scripts/run.py only for:

  • deterministic assembly of timeline bullets from paper_notes + bibkeys
  • table generation from outline + mapping
  • figure spec skeleton generation using assets/figure_templates.yaml

Do not treat run.py as the place for:

  • hardcoded figure descriptions or narratives
  • milestone selection heuristics that should be inspectable from references

Tables are handled by dedicated table skills:

  • table-schema -> outline/table_schema.md
  • table-filler -> outline/tables_index.md (internal index)
  • appendix-table-writer -> outline/tables_appendix.md (reader-facing Appendix tables)

Inputs

  • outline/outline.yml
  • outline/claim_evidence_matrix.md
  • papers/paper_notes.jsonl
  • citations/ref.bib

Outputs

  • outline/timeline.md
  • outline/figures.md

Workflow inputs (explicit)

  • Use outline/outline.yml + outline/claim_evidence_matrix.md to decide what to visualize.
  • Use papers/paper_notes.jsonl for year/milestone candidates.
  • Use only citation keys from citations/ref.bib.

Workflow (heuristic)

1) Read the outline + claim-evidence matrix and pick recurring comparison axes.

2) Timeline (outline/timeline.md):

  • Write year -> key milestone bullets (aim for breadth and citations).

3) Figures (outline/figures.md):

  • Write 2-4 figure specs that a human could draw:
  • purpose (what insight this figure communicates)
  • required elements (boxes/arrows/axes)
  • what papers support each element (cite keys)

4) Use only citation keys present in citations/ref.bib.

Quality checklist

  • [ ] No TODO and no <!-- SCAFFOLD ... --> markers remain in the outputs.
  • [ ] outline/timeline.md contains >=8 year bullets and each bullet has >=1 citation marker [@...].
  • [ ] outline/figures.md contains >=2 figure specs and each mentions at least one supporting citation.

Helper script (optional)

Quick Start

  • uv run python .codex/skills/survey-visuals/scripts/run.py --help
  • uv run python .codex/skills/survey-visuals/scripts/run.py --workspace <workspace>

All Options

  • --workspace <workspace> (required)
  • --unit-id <id> (optional; used only for runner bookkeeping)
  • --inputs <a;b;c> (optional; defaults to the four Inputs listed above)
  • --outputs <timeline_rel;figures_rel> (optional; defaults to outline/timeline.md;outline/figures.md)
  • --checkpoint <C#> (optional; ignored by the helper)

Examples

  • Generate timeline + figures with defaults:

uv run python .codex/skills/survey-visuals/scripts/run.py --workspace <workspace>

  • Generate to custom output paths:

uv run python .codex/skills/survey-visuals/scripts/run.py --workspace <workspace> --outputs outline/timeline.md;outline/figures.md

Notes

  • The helper is intentionally minimal and never overwrites non-placeholder artifacts.
  • In strict mode it blocks only if placeholder markers remain (and if minimum timeline/figure requirements are not met).

Troubleshooting

Issue: timeline is thin or citation-free

Fix:

  • Prefer fewer, higher-signal milestones, but ensure each bullet has >=1 [@...].
  • Route upstream if notes are thin: strengthen paper-notes / evidence-draft rather than padding.

Issue: figure specs read like prose

Fix:

  • Keep specs as draw-instructions: purpose + elements + what each element is supported by (cite keys).
  • Move narrative explanation into the main text; this file should stay non-prose.

How to use it

Copy the folder

Take willoscar/survey-visuals from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.