SciStudio

Base identity for the SciStudio embedded agent. Lists the 6 task skills available and when to invoke each. Loaded once at session start; task skills load on demand when the user turn matches their trigger description.

Drive a live end-to-end SciStudio session — start the backend (and Vite if needed), open Chrome via Chrome MCP, walk the scenario file the user points at, capture screenshots/console/network as evidence, and write a pass/fail verdict back into that file. Trigger whenever the user says "跑 e2e", "do an e2e", "run e2e on PR

Use when the user wants to design a new workflow, choose block types, wire edges between blocks, or convert a verbal pipeline description into a valid workflow YAML. NOT for debugging existing runs (use scistudio-debug-run) or for writing custom blocks (use scistudio-write-block).

Use when the user wants to look at intermediate or output data — preview a slice of an image, peek at the first rows of a DataFrame, check the shape/dtype of an array, or trace where a data ref came from. NOT for debugging failed runs (use scistudio-debug-run).

Use when the user wants to author a NEW BLOCK FILE — Python source code (a class subclassing Block / ProcessBlock / IOBlock / AppBlock / CodeBlock with typed ports and a config_schema) that goes in ``<project>/blocks/<name>.py``. ALWAYS check if an existing block satisfies the contract first (call list_blocks first and reuse a match). NOT for ADDING AN EXISTING BLOCK TYPE AS A NODE in a workflow YAML — that's scistudio-build-workflow ("add an imaging.threshold node to my workflow"). NOT for editing an existing block's config (update_block_config). NOT for writing the workflow YAML itself (scistudio-build-workflow).

Use when a workflow run has failed, is stuck, or produced unexpected output, and you need to diagnose the cause. Covers run-status inspection, block-log retrieval, lineage navigation, common error signatures, and the finish_ai_block contract for AI block PTYs. NOT for designing new workflows (use scistudio-build-workflow).

Use when the user asks a question ABOUT SciStudio or about this project — how a feature works, what a block/type/contract is, what blocks are installed, where docs/data live, project name / metadata, recent workflows. The Q&A skill. NOT for designing or debugging workflows (scistudio-build-workflow / scistudio-debug-run) or authoring code (scistudio-write-block / scistudio-write-plot).

Use when the user wants a QUICK PREVIEW PLOT: a matplotlib, seaborn, base R, or ggplot2 figure drawn from a block OUTPUT collection and shown in the preview panel (e.g. "scatter the measurements from my segment-cells node", "show me a histogram of that output"). A plot job lives in ``<project>/plots/<plot_id>/`` and renders a display-only artifact. NOT for authoring a reusable processing BLOCK: that is scistudio-write-block (a Python class under ``blocks/<name>.py`` that becomes a workflow node). NOT for adding a node to a workflow YAML: that is scistudio-build-workflow. A plot job NEVER becomes a workflow DAG node, NEVER edits workflow YAML, and NEVER produces downstream data or lineage.

Use when acting as the SciStudio live implementer persona for owner-directed live implementation sessions under the guided task kind.

Use when acting as the SciStudio test_engineer persona for test architecture, test design, test additions, runtime validation, e2e evidence, and test-only artifacts.

Use when acting as the SciStudio ADR author persona for ADRs, addenda, specs, governance text, document standards, schemas, and architecture-facing decisions.

Use when acting as the SciStudio audit reviewer persona for reviews, audits, CI investigation, drift detection, conformance checks, and readiness assessment.

Use when acting as the SciStudio implementer persona for scoped code, tests, tool wiring, implementation-linked docs, and gate-record evidence.

Use when acting as the SciStudio manager persona for planning, coordination, dispatch, hotfix ownership, integration testing, e2e testing, PR readiness, and durable status reporting.

Use when running a live end-to-end SciStudio session — driving the GUI in Chrome via Chrome MCP, verifying a PR before merge, reproducing a hotfix bug live, or any time the user says "跑 e2e", "do an e2e", "verify in Chrome", "run e2e on PR

Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.

Generate a custom checklist for the current feature based on user requirements.

Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.

Execute the implementation plan by processing and executing all tasks defined in tasks.md

Execute the implementation planning workflow using the plan template to generate design artifacts.

Create or update the feature specification from a natural language feature description.

Generate an actionable, dependency-ordered tasks.md for the feature based on available design artifacts.

Convert existing tasks into actionable, dependency-ordered GitHub issues for the feature based on available design artifacts.