joshua-zyy/academic-paper-writer

Academic Figure

CS/AI/ML academic-figure router. Experimental data figures are default delivered as editable Python/matplotlib SVGs and pass through the CS/AI/ML chart design gate. Model framework diagrams, architecture diagrams, overview figures, and complex mechanism diagrams are out of scope for automatic rendering by this skill; if the paper requires such figures, only human-drawing requirements, evidence checklists, caption drafts, or figure blueprint notes may be produced.

Router Protocol

1. Read manifest.yaml. It declares always_load files, axes, and references.on_demand.
2. Read every file listed under always_load. These are the skill's binding rules — not reference material.
3. Apply the loaded material as constraints:
   • stance.md defines Python-only plotting, figure contract, visual policy, and scope.
   • red-lines.md defines absolute prohibitions. Do not negotiate these.
   • output-contract.md defines deliverables per mode.
   • anti-patterns.md defines known failure modes and their correct alternatives.
4. Select exactly one mode from the manifest. If ambiguous, ask one concise clarification only when data source or target use is missing. Requests for model framework, architecture, overview, or mechanism diagrams use figure-blueprint only to produce manual_figure_needed notes; never render images, SVG, or prompts.
5. Echo the selected mode to the user before executing.
6. Reach for references/ only when the manifest's references.on_demand condition is satisfied.

Modes

Mode	Use when
chart-from-data	Data files or numeric results, needs publication plot with CS/AI/ML chart design gate
figure-blueprint	Wants figure suggestions for a paper section
figure-audit	Existing figure reviewed for publication readiness
figure-revision	Existing figure needs revision

Agent Dispatch

agents/figure_agent.md is dispatched by the orchestrator at Step 6.4. The agent returns figure artifacts, scripts, SVG paths, and reports; it must not independently edit project source code or experimental data.

Completion Criteria

• chart-from-data: Figure Contract, CS/AI/ML chart design gate, Python script, source data, editable SVG, QA report — all pass.
• figure-audit: Every QA item has pass/fail status and concrete remediation.
• figure-blueprint: Every suggested figure maps to a paper claim and data/evidence source.
• figure-revision: Revised artifact or instructions, QA report, unchanged evidence traceability.