SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillslight-tool-selection

light-tool-selection

Coding & development

View repository

Updated 10-16 00:00

About this skill

Tool Selection and Multi-Tool Collaboration

Principles (Always Active)

Do not use tools for the sake of using tools. For each task, first ask:What is the most efficient, most stable, most professional, and most reproducible way? Prefer reproducible (code/script) > one-off manual work; prefer the stack already in the project > introducing new dependencies.

Task → Recommended Tool Mapping

  • Literature Search：arXiv/OpenAlex/Semantic Scholar/Crossref API、agent-browser/browser-use、Exa/Parallel Web。
  • Browser Automation: For CLI-oriented/stable cases, use agent-browser (native Rust CLI, direct CDP connection, accessibility-tree snapshot +@eN citation location, keep-alive between commands, observable panel on port 4848; first agent-browser skills get core to discover usage); for Python-oriented integration, use browser-use (Agent(task,llm,browser)+await agent.run(), based on Playwright,@tools.action add custom tools,use_cloud=True use stealth cloud browser).Submission system web operations(form filling, uploading, status checking in OpenReview/Editorial Manager, etc.) can install agent-browser-type skills (about 438K installs, last_checked 2026-06),Light does not build its own browser automation, introduce external skills as needed and evaluate source security.
  • Data Processing: pandas (small to medium) / polars (fast) / DuckDB·polars streaming·dask (beyond memory); quality: ydata-profiling/Deepchecks/Great Expectations.
  • Statistics/Scientific Computing: Python (statsmodels/scipy/sklearn), R (advanced statistics/plotting), MATLAB (signal/control/numerical).
  • Plotting: matplotlib/seaborn/plotly/altair, ggplot2, Origin (journal curves), TikZ/Graphviz/Mermaid (framework flowcharts),Draw.io(framework/system/architecture diagrams as code, has official MCP, XML can be version-controlled), Illustrator/Inkscape (fine retouching), Visio.
  • 3D Visualization/Rendering: Blender (open source, procedural modeling and rendering, not AI image generation; for 3D scientific visualization use plugins such as Molecular Nodes/SciBlend, 3D rendering for roadshows; has community/official MCP, requires local installation). Pair with m09 figure-planning / m16 slides.
  • Typesetting: LaTeX (latexmk/TinyTeX/TeX Live), Word (python-docx), Pandoc (mutual conversion).
  • Document Processing：PDF/DOCX/PPTX/XLSX skill、MarkItDown、unstructured.io、Apache Tika。
  • PPT：python-pptx/PptxGenJS、Marp、reveal.js、Beamer、PowerPoint/Canva/Gamma。
  • Frontend: Next.js/React, shadcn/ui, Tailwind, ECharts/D3; when there is a design mockup, use Figma MCP(read mockup → IDE generates code, Remote server is free and supports read/write), pair with a05 frontend-design.
  • Backend：FastAPI/Django/Spring Boot、Postgres/Redis、Docker。
  • API Calls: If there is an OpenAPI description, call deterministically according to paths→operationId+parameter schema+securitySchemes and parse according to the response schema; do not guess from prose documentation (for field details, see references.md).
  • Cloud Compute: For Python serverless/GPU training/sandboxes, use Modal (define the environment in code with @app.function(gpu=..., image=...) , billed by the second; for commands and primitives, see references.md).
  • Versioning/Experiments：Git、DVC、MLflow、W&B、Hydra、Snakemake。
  • Automation/CI: Repository event triggering uses GitHub Actions (test-on-push, scheduled runs, auto-building paper PDFs/figures, release); local data dependency orchestration uses Snakemake/Make — the two are complementary (Actions handles "what event triggers", Snakemake handles "data dependencies between steps", see references.md).
  • Reference management：Zotero/pyzotero、JabRef、Better BibTeX。
  • Environment: Pure Python prefers uv; compiled scientific libraries/CUDA/cross-language use conda (mamba/miniforge); existing Poetry projects continue with Poetry.

Key decision points

  • Data scale → library choice (pandas vs polars vs dask).
  • One-off vs reproducible → manual vs script.
  • Output destination → determines format (vector graphics for submission, png for presentation).
  • Team/reproducibility needs → whether to adopt version and experiment management.
  • Stability > novelty: choose mature, actively maintained tools (CONVENTIONS dependency safety).

Withnumeric thresholdsThe complete decision matrix (data <100MB pandas / 2-50GB polars / >TB Spark; submission vector vs presentation png≥150dpi; skill installs ≥1K, etc.) is in references/decision_matrix.md, load as needed.

Auto-detect project tech stack

When entering an existing project, or when the user asks "what tools should this project use", run the detection script first rather than manually going through a checklist:

python scripts/detect_stack.py <项目目录>        # 读 package.json/pyproject.toml/requirements.txt/environment.yml
python scripts/detect_stack.py <项目目录> --json # 机器可读
python scripts/detect_stack.py --self-test       # 无项目时合成清单自检

The script identifies dependencies → gives selection recommendations by category (only for those matching built-in rules; non-matches are marked no signal, not fabricated), and based on lock files (uv.lock/poetry.lock/environment.yml/Dockerfile) gives environment/reproducibility recommendations. Report according to assets/stack_detection_report.md.

Python environment selection

For the decision table choosing uv / conda / Poetry / Docker by dependency nature (scenario→tool→key commands, mutually exclusive iron rules), see the authoritative version in references/decision_matrix.md section 5; for complete per-tool commands/parameters/pitfalls see references.md (uv/Conda/Poetry/Docker sections). In one sentence: pure Python/PyPI prefers uv, compiled scientific libraries/CUDA/cross-language use conda(mamba/miniforge), existing Poetry projects continue with Poetry, portable service reproducibility uses Docker; all three rely on lock files to ensure reproducibility,do not mix pip+conda in the same environment。

Multi-tool collaboration

A single task often spans tools: e.g., "experiment→W&B logging→pandas aggregation→seaborn plotting→LaTeX typesetting→Zotero citation". Plan the data flow format between tools well (CSV/JSON/Parquet) to reduce manual transfer.

Tool discovery

When new capabilities are needed, proceed in order:

  1. First check existing skills: first look at the skills.sh leaderboard (by total installs; top ones like vercel-labs/agent-skills, anthropics/skills are all 100K+, last_checked 2026-06), then search npx skills find [关键词] .For the authoritative version of quality numeric thresholds (install count/GitHub star/source credibility), see references/decision_matrix.md section 7. List for the user name/purpose/install count/source/link/install command, and after consent npx skills add <owner/repo@skill> -g -y(-g global, -y skip confirmation); if no match, be honest, switch to general capabilities, and suggest npx skills init <name> to build your own. (Install counts change over time; when citing specific numbers always include last_checked.)
  2. For external data/tools: check available MCP server(reference picks Everything/Fetch/Filesystem/Git/Memory/Sequential-Thinking/Time; for TS use npx -y @modelcontextprotocol/server-*, for Python use uvx mcp-server-*, on Windows npx needs cmd /c wrapper; registry.modelcontextprotocol.io has more). First tools/list/resources/list See what is exposed before calling.For research design/plotting/3D/computation categories, recommended MCPs have been selected(suggest on demand when routing to the corresponding skill): Figma (read design mockups → frontend, a05), Canva (roadshow PPT/posters, m16/m17), Draw.io (framework/system/architecture diagrams diagram-as-code, m09/a04/m15), Blender (3D scientific visualization/roadshow rendering, m09/m16), MATLAB (signal/control/numerical/Simulink)——the single source of truth for capabilities/cost/stars is in README recommended MCP table, this section does not restate the numbers (same convention as the OpenAlex key).
  3. Building your own skill: use the skill-creator route——SKILL.md(frontmatter only name/description)+scripts/+references/+assets/, three-level progressive disclosure, the description enumerates trigger scenarios to prevent under-triggering, and the body uses imperative sentences to explain "why". Evaluation loop: real test prompts → with-skill vs baseline quantitative comparison → read transcripts for generalization, don't overfit → extract repeated work into scripts/. (For skill-creator/Autoskill details see references.md.) Introducing any third-party skill/MCP server is equivalent to authorizing external instructions and code; evaluate the source and security first. See references.md for details.

Resource list (bundled)

  • scripts/detect_stack.py(tech stack detection,--self-test offline self-check),references/decision_matrix.md(decision matrix with numerical thresholds),references.md(per-tool endpoints/commands/pitfalls).
  • Presentation templates assets/：stack_detection_report.md(detection report),skill_discovery_report.md(skill discovery, single/multiple candidates/no-match fallback),mcp_discovery_report.md(MCP discovery, including tools/list, resources/list probe checklist). Both discovery templates include 2026-06 tested entry points and HTTP codes.

Connection

Provides the "what to use to do it" judgment for all skills; works in concert with a06 (directory), a02 (version tools), a03 (code stack).

For per-tool hard information (real endpoints/commands/parameters/pitfalls) see references.md in the same directory.