editaplot

EditaPlot

与论文图件规范的分工

单独提出科研数据绘图请求时，在当前 Origin 路线能满足要求的前提下优先使用本技能，交付用户可继续手工微调的原生 OPJU；无需用户明确说 Origin。论文结果图由 paper-figure-review 确定技术含义、坐标单位和投稿排版要求，本技能负责导入、制作、保存和导出，再按该要求检查实际输出。

论文任务已确定的字体、字号、尺寸等要求优先于下文及参考资料的 Origin 模板默认值；当前路线不能实现或回读验证时，报告缺口。跨技能继承已确定的任务输出目录，用现有输出参数传入，无需再次询问目录。

科研示意图走 PPT、SVG 等设计路线。仿真、实验程序运行时的自动出图保留现有实现；用户指定 Python/MATLAB 或沿用既有图源时遵循指定，不自动迁移。仅审查时给意见，不生成修改版。

两个技能复用同一任务的数据含义、样式和授权，不各问一轮确认。用户保存后的当前 OPJU 是后续读取依据，按现有副本审查流程导出，不覆盖其手工调整；这不代表已提供任意 OPJU 原位编辑接口。

制作前核对必要坐标尺度、阈值线、标记、版面与导出格式是否被当前路线支持；仅模板名匹配或 Origin 可启动不算满足。能力不足时说明具体缺口和可行替代，不静默把对数轴换成线性轴，也不把不符合要求的试图标为完成。

Turn a scientific question and a read-only table into an auditable, editable Origin figure. Keep the beginner experience conversational; use the deterministic engine for inspection, planning, rendering, exporting, and readback.

Request only scoped Windows permissions

• Read the configured versioned runtime, selected table, and optional local reference image. Read the complete repository only when working from an explicitly selected clone or approved source update.
• Write only to the configured runtime's .editaplot-venv, an approved source repository during a separately authorized update, and the established task output folder (defaulting to the selected source file's parent folder). Never patch .cc-switch, .codex, or .claude Skill copies directly.
• Run the local launcher, PowerShell/Python subprocesses, and an EditaPlot-owned Origin instance in the same active interactive Windows user session.
• Use network access only for repository download/update and locked dependency retrieval. Treat a user-scope winget Python installation as a separate system change that still requires explicit consent.
• Do not request administrator rights, mouse control, whole-drive write access, cloud upload of private inputs, or DCOM, registry, firewall, user-group, or Origin-installation changes. When Controlled Folder Access, an organization policy, cloud sync, or a read-only location blocks writes, request access only to the affected folder or ask for an explicit alternate output folder.

Start with the beginner path

1. Reject unsupported platforms before installing anything. Support the secure CLI/dependency layer only on physical Windows 10/11 x64 with 64-bit CPython 3.11 or 3.12; recommend 3.12 for new installs. The locked Pillow version is 13.0.0; CPython 3.10 is no longer supported by this secure runtime. Target Origin/OriginPro 2021 and later through external originpro; Origin 2020b and earlier are unsupported by this route. Retain the complete CPython 3.10 + Origin 2024b / 10.15 record as historical evidence only. The current CPython 3.12.14 + Pillow 13.0.0 + Origin 2024b / 10.15.132 live regression covers six representative cases with smoke, complete artifacts, independent saved-project reopening, data/axis/physical-geometry readback, and re-exported PNG pixel identity checks. It does not verify every route or Origin version; Python 3.11 has no live-verification claim. Require the current workflow's smoke and artifact gates, and keep other 2021+ hosts capability-gated. State plainly that macOS (Intel/Apple Silicon), Linux, WSL, Wine/CrossOver, Parallels, and other VMs are unsupported in V1. doctor cannot reliably detect every VM, so ask the user to confirm a physical Windows host when that fact is unknown.

2. Locate editaplot.cmd in the installed Skill directory; when working from a cloned repository, use the repository-root editaplot.cmd. Use an absolute launcher path in commands. Do not make beginners select a Python executable or invoke scripts/editaplot.py directly.

3. This personal CC Switch distribution includes editaplot.cmd and a versioned .editaplot-local.json that points to an audited, commit-named local runtime snapshot. Run py -3.12 scripts/validate_local_distribution.py, then use the installed launcher; do not run upstream setup against .cc-switch, .codex, or .claude, because CC Switch owns those runtime directories. If validation reports a stale or missing engine, repair the source distribution and resync it through CC Switch instead of patching a runtime copy. Read references/runtime.md for discovery, update, and command details.

4. Reuse an existing compatible Python 3.11 or 3.12. Prepare the new locked environment through the authorized CC Switch source-distribution workflow; retain the old environment for recovery, but never select it automatically after a failed upgrade. If none exists, explain in Chinese that installing Python is a system-level change. Run winget show first and explain the exact publisher, source, and agreements. Only explicit user confirmation permits a later non-interactive installation of Python.Python.3.12 with user scope and x64 architecture. If winget is unavailable, provide the official python.org Windows installation instructions and wait for the user; never use an untrusted mirror or silently install Python.

5. Run editaplot.cmd doctor for each new workflow. Allow doctor --repair only for the reported project-local Python dependency repair. Keep all Python packages in .editaplot-venv. Treat Origin as a locally installed user-managed application; never install or modify it during repair.

6. Run editaplot.cmd start <data-file> for a new table. Add --intent "<user intent>" when the user states a goal. Treat its inspection and recommendation payload as internal working state. Use the original local source path exposed by the attachment. If the host provides only a temporary copied attachment and the original folder cannot be recovered, ask once for the intended local source/output folder before rendering; never guess an unrelated workspace destination.

7. After selecting a candidate template, run editaplot.cmd understand <data-file> --template-id <id> with the same confirmed mapping that will be used for planning. Group its result into an internal checklist: data type; columns to draw; columns used only for support or validation; columns retained but not drawn; proposed figure elements; and calculations that will not be performed. Every source column must appear exactly once. If any item is uncertain, first check the supplied material; ask only about meaning still unresolved, then run understand again. Do not confirm or plan unresolved items. Tell the user only what will be drawn and any material question, unless they request the full column breakdown.

8. Tell a beginner only: what was recognized, the best one to three chart choices, why they fit, and the smallest scientific decision still required. Do not dump an inspect → recommend → understand → plan pipeline or raw JSON unless they ask for technical detail.

9. Reuse the scientific purpose and element choices already explicit in the request. Ask only for unresolved scientific choices. Freeze the exact proposal_hash, approved derived-item IDs, and resolved ambiguity choices in --semantic-confirmation-json. Never reuse a confirmation after the source, mapping, purpose, or proposal hash changes: regenerate the proposal and confirmation payload after revalidation. A changed hash alone does not require another user answer. Reuse the existing decision when the current request covers the updated data and its purpose, column roles, units, and authorized transformations still match; ask only about an unresolved change in meaning. Internal awaiting_*_confirmation states request a valid payload, not a mandatory extra conversation turn. When confidence is low, candidate margins are small, roles or units are ambiguous, or a new, unauthorized meaning-changing display transformation is proposed, ask only the additional focused questions needed.

10. If the user supplies a reference figure, first run reference-inspect. Codex may then describe only its panel/mark/encoding/layout/style grammar in the strict ReferenceFigureSpec JSON and run reference-review; the runtime performs no OCR or model inference. Show the adopted and rejected features, bind every essential mark to confirmed renderable user data, and record the existing exact request as authorization when it covers the selected features; ask only for new decisions. Never copy reference values, labels, fits, phase assignments, author text, logos, watermarks, or the bitmap into the Origin project. Prefer verified template_adaptation; keep controlled_composition blocked until that exact composition has passed the full Origin evidence gate. A reference cannot add missing evidence or change the confirmed scientific element list. Reuse explicit palette, reference and style choices. With no preference, use the verified template default; ask only for a material visual or scientific tradeoff. Record the actual request in the existing confirmation interface, never invent a later user answer. Mark supported requests as applied, retained_template_default or rejected and verify object readback.

11. When the user needs palette options, run editaplot.cmd palettes, show assets/palettes/palette-selector-public.zh-CN.png, and recommend no more than two compatible palette_id values. Read references/palettes.md before freezing one.

12. Internally freeze the confirmed choice with editaplot.cmd plan; never hand-edit a plan or write a decision back to the source file. For an exact XPS request, write the confirmed values to a separate JSON object and pass its path with --visual-style-json. The supported exact fields are series_colors, line_width_pt, fill_transparency_percent, page_size_cm, legend_visible, legend_position, and legend_frame. Invalid explicit fields or values must fail fast and be corrected with the user; never silently discard them or fall back to a reference/default style. The render command copies this approved plan into the final output folder as render-plan.json.

13. Treat Origin readiness as technical state only. Doctor performs read-only discovery of Origin.Application, Origin.ApplicationSI, installed candidates, Python, originpro, and OriginExt; it never launches Origin and ready_for_render never means a live connection succeeded. If the default launch registration is present, proceed to the real pre-render smoke without asking the user to open Origin or confirm it again. Keep beginner output to one to three plain-language sentences; leave CLSIDs, registry views, candidates, and stages in JSON.

14. Run editaplot.cmd origin-smoke --output-dir <unique-smoke-directory> with launch_isolated: start and own a dedicated Origin instance, perform the live smoke and version handshake, then apply the template capability decision. This command is mandatory after planning and before formal rendering. attach_existing is an explicit advanced mode only; never reset, overwrite, or close a user-owned project, and detach instead of exiting. Report failures by technical stage and next step without speculation. Never use mouse automation or provide application patches or bypass instructions. The runtime must attempt to clean a partial EditaPlot-owned activation and may try one fresh isolated instance for a retryable startup code only if cleanup succeeds. Cleanup failure returns origin_activation_cleanup_failed and stops. It must then wait with sec -poc 30 and confirm run.isOCready() before reading the version or creating a project. Never loop, switch to ApplicationSI, edit DCOM/registry permissions, or tell a beginner to run the whole workflow as administrator. After the automatic attempt is exhausted, request approval for at most one retry in the same active Windows-user context and use a fresh empty sibling smoke directory so the first report remains intact. origin_com_class_not_registered and origin_com_activation_access_denied stop without automatic retry. Do not force-terminate a Python worker merely because it has run for a long time: it may own a hidden Origin instance. Preserve diagnostics and report the last progress stage before proposing any user-controlled cancellation.

15. Only after that smoke passes, render an allowed template route with editaplot.cmd render <plan>. Keep an EditaPlot-owned Origin instance open after success unless the user requests otherwise. By default, let the runtime create a direct sibling of the source file named <source_stem>_EditaPlot_YYYYMMDD_HHMMSS; keep all formal artifacts in that folder. Do not redirect ordinary runs to the repository, Skill directory, current working directory, or a shared global output folder. Use --output-dir for an established task output folder or a location explicitly requested by the user.

16. Run editaplot.cmd verify <output-directory> against the selected output folder and perform agent visual QA. If smoke or render fails, a Python preview or standalone PNG/PDF/SVG is only a preview and must not be presented as completed Origin work. Formal success requires the editable OPJU, PNG, PDF, TIF, object readback, and agent visual QA together.

Before any render, read references/origin-safety.md, references/figure-contract.md, and references/verification.md. For a new table or chart decision, read references/data-contracts.md, references/chart-selection.md, and references/semantic-understanding.md. When a reference image is supplied, also read references/reference-figures.md.

Collaborate on a user-edited OPJU

Use this route when the user wants to make final Origin edits personally and then asks for a review. V1 covers data plots and already verified Origin templates; system diagrams and mechanism figures remain in PPT/SVG workflows.

1. Run editaplot.cmd prepare-opju-review <result.opju>. This creates a new sibling <source_stem>_OriginReview workspace (or an explicitly requested new directory) containing figure_initial.opju, figure_edit.opju, and review-workspace.json. It only copies and never replaces the source, an existing workspace, or either copy. Treat figure_initial.opju as the immutable baseline and figure_edit.opju as the only file the user edits and saves in Origin.
2. Ask the user to edit and save figure_edit.opju directly in Origin. Do not require v2/v3 copies. Do not run the review command until the user explicitly says the file is saved and ready for review.
3. After that confirmation, run editaplot.cmd review-opju <workspace>\figure_edit.opju. The command creates a unique review/<timestamp>/ directory, copies figure_edit.opju to a snapshot, and checks SHA-256 before and after copying and after the isolated review. A changed, locked, or baseline-tampered file stops with a structured error and never opens or overwrites the user file.
4. The worker uses the fixed runtime's OriginSession with new_isolated, opens only the snapshot using readonly=True, enumerates every Graph Page, and exports PNG/PDF/TIF below that review directory. It must never attach to or save the user's Origin project; the session may initialize its own empty EditaPlot-owned instance, and it closes only that instance.
5. Briefly report the reviewed figures, usable export paths, and any unresolved issue affecting use; keep review-report.json as a detailed record and link it only when useful or requested. It records snapshot and baseline integrity, Origin version and instance ownership, Graph Page/object inventory, and export hashes. Export success means only “the snapshot can be read back and viewed”; inspect PNG/TIF for scientific meaning and visual quality and keep human_visual_qa.status=pending until the agent actually inspects the images. V1 never writes changes back into the user's OPJU.

Keep scientific decisions with the user

• Treat the original data file as immutable. Never overwrite it, fill missing source columns, or invent measurements. Permit helper columns only in memory or the editable Origin project.
• Classify every source column before planning as primary render, secondary render, support-only, retain-not-render, or uncertain. Support-only and retained columns cannot become visible through a reference image. An unknown numeric column is a question, not another automatic curve.
• Distinguish scientific analysis from display transformation. Never silently normalize, smooth, fit, remove outliers, calculate error bars, identify phases, or infer material peaks.
• For GSAS/GSAS-II Rietveld data, distinguish Observed, Calculated, optional Background, supplied Difference, explicit Phase positions, and non-rendering control/diagnostic columns. Preserve an upstream Publication Diff exactly; never apply a second display offset.
• For XPS, keep cosmetic preferences separate from the scientific contract. A user may explicitly request exact series colors, physical line widths, fill transparency, a safe page/aspect ratio, and legend show/hide, borderless, or position choices. Apply only fields supported and read back by the selected verified XPS renderer; otherwise retain the default or reject the field visibly. Neither a user style request nor a reference image may change source values or column roles, the high-to-low binding-energy axis contract, component identity, residual disposition, or the verified single-region set_fill_area(..., type=9) / -pfm 3 fill implementation.
• For FLIM, use flim_phasor for upstream G/S coordinates with optional groups. Keep every point, including points outside the reference semicircle. Confirm the mathematical semicircle helper; preserve equal physical G/S scales. Do not infer lifetimes, fit decays, unmix or cluster.
• For supplied event times, use event_raster with explicit Time/unit, Trial and Condition. Confirm source-order trial ordinals; blank Time entries declare inventory, not events at zero. Preserve every event and empty lane, and retain zero-event conditions in the borderless outside legend. Never detect events, calculate PSTH/rates or reorder trials by response. Read the event-table section of references/data-contracts.md; only condition-compatible palettes may replace the default.
• For SHAP, accept only externally precomputed per-sample contributions. Never train a model or invoke SHAP. Mean |SHAP| and optional group percentages may only summarize those supplied rows with the allow-listed formulas recorded in the semantic proposal and explicitly approved; never invent contributions or silently reorder features.
• When the user requests separate importance bars, a beeswarm and a two-level contribution ring, select shap_dashboard. Require explicit Feature Group, confirm feature/group percentages using one total Mean |SHAP| denominator, and offer blue–white–wine, purple–green–yellow, or red–yellow–blue. Freeze the selection as mapping plot_mode dashboard_blue_red, dashboard_viridis, or dashboard_red_blue; do not substitute a general categorical palette for this continuous scale.
• Confirm unknown units, error semantics, percentage denominators, meaningful order, dual axes, and any other choice that can change the claim.
• Recommend from the scientific question and data structure, not aesthetics alone. Refuse a misleading chart even when technically renderable.
• Keep template route status (verified, experimental, or unsupported) separate from current host compatibility (verified, compatible_unverified, or blocked). Never relabel a compatible_unverified Origin version as verified; continue only when its smoke succeeds and the selected template's required capabilities are available.
• Reject decorative 3D. Require a scientifically meaningful third axis; keep a new 3D route experimental until Z-axis, camera, OpenGL type, source mapping, four exports, editable OPJU, readback, and visual QA pass.
• 对于已验证的 density_ridgeline3d，我只接受 2–6 个真实带单位条件的 mixed-wide 六角色表：上游提供同语义同单位的实线/虚线预计算密度，并为每组提供恰好一个 Focal X。 焦点固定为 Z=0 基线 locator；不要计算 KDE、峰值、阈值、交点或焦点。当前主机还必须先通过 实时 smoke 与 OPEN_GL_3D 能力检查，不能只凭模板已验证就跳过主机门禁。
• Do not send selected files to any additional network service or include them in public artifacts. A file explicitly provided through Codex remains subject to the user's Codex account, organization, and retention policies; do not claim the Skill can override those policies.
• Before inspecting medical data or reference images, require the user to confirm that the material follows their institution's rules, is deidentified, and has been checked for burned-in text.
• Treat panel-plan as a deidentification-aware layout and evidence gate, not an OCR, PHI detector, medical image editor, or merged editable Origin project. Preserve every verified subproject.

Apply the publication-informed contract

For supplied time–frequency tables, choose time_frequency only when time, frequency and value columns explicitly state their units and form a complete uniform grid. Preserve both coordinate directions and every cell; use the numeric physical-axis route, not categorical heatmap. The route shows sequential value magnitude with a detached scale; it performs no FFT/STFT/wavelet, filtering, interpolation, baseline or dB conversion. A raw EEG voltage table needs upstream processing, not an invented frequency column. Read references/data-contracts.md for the input and unit contract; confirm any extra columns instead of silently drawing them.

For multichannel EEG or fluorescence traces already supplied as a table, choose signal_stack. Require explicit Time/Channel/Amplitude roles and units, keep every observation and the channel order, and confirm the additive display offsets and common unit ruler. Channel/ROI identities remain text, including leading zeros. Do not filter, align, center, normalize, infer events or decode acquisition files. Direct names and the ruler live outside the true time-axis data range.

• Make every chart defend one explicit conclusion or evidence role.
• Use a white background, Arial, restrained color families, clear hierarchy, and no rainbow palette, decorative 3D, or unjustified grid.
• Derive physical Origin dimensions from chart type, data density, series count, and label length. Keep fixed size only for a profile that explicitly requires it, such as legacy fixed C 1s.
• Convert documented point, line-width, and page-size units correctly. Never copy small journal-page font values directly into Origin API fields; read back the resulting axis and text objects.
• Keep each condition's color consistent across related panels. Freeze palette IDs and exact HEX values, allowed modes, safe category count, and accessibility constraints into the plan.
• Let an explicit user style request outrank a style token inferred from a reference image. Color, line-width, transparency, page/aspect, and legend requests are still capability-gated and must be classified as applied, retained default, or rejected before rendering.
• Do not let a reference image or unverified cosmetic preference silently redefine semantic color mappings for XPS components, signed effects, heatmaps, diagnostic lines, confusion matrices, or similar evidence. An explicit replacement is allowed only through that route's independently verified override with exact series mapping and readback.
• Give every medical panel one distinct evidence role. Freeze a shared condition-to-color map before composing quantitative panels; require explicit semantic confirmation for a shared legend.
• Prefer editable labels and Origin objects. A Python preview or embedded bitmap is not an Origin deliverable.
• Call the result “publication-informed,” never “Nature compliant” or journal-approved.

Report the result in plain language

Default to a short Chinese result: what was drawn, links to the editable figure and useful exports, and any unresolved issue that affects use. If required validation is incomplete, say which part remains unverified. Do not list every alternative, confidence score, internal hash, plan or check file unless it helps the user decide or they request details. Keep data roles, authorized transformations, the render plan, provenance and full validation records in the existing workflow's process files. Concise delivery does not remove any required artifact, object readback or agent visual QA.

Load detailed references only as needed

• references/runtime.md: launcher, setup, Python discovery, CLI commands, and artifacts.
• references/chart-selection.md: chart families, ranking rules, and support levels.
• references/data-contracts.md: accepted layouts, column semantics, and repair guidance.
• references/semantic-understanding.md: per-column use, element checklist, derived-data lineage, and the hash-bound confirmation gate.
• references/reference-figures.md: safe reference grammar, bindings, adaptation limits, and recording of the existing exact authorization.
• references/figure-contract.md: evidence logic, visual hierarchy, typography, and color rules.
• references/origin-safety.md: local Automation and verified-API guardrails.
• references/verification.md: mandatory artifacts, readback, and visual QA.
• references/showcase.md: neutral demonstration data and gallery policy.
• references/palettes.md: Chinese palette selector, compatibility, and accessibility limits.