5 releases 5 tags

RSS feed

• v0.2.0 7c8b0ece9f

  Compare

  v0.2.0 — Data Scientist + EPUB Skills Stable

  jasper released this 2026-05-23 18:10:52 -04:00 | 13 commits to main since this release

  v0.2.0 — Two New Skills, 50 Files, 12,792 Lines

  This release adds two major new class-level skills — data-scientist and epub —
  plus significant expansions to both. 50 files changed, +12,792 lines.

  New Skill: data-scientist (data-scientist/)

  PhD-level data science expertise for AI agents. Eight competency domains, from
  mathematical foundations through experimental design, causal inference, Bayesian
  analysis, ML, MLOps, and research leadership.

  • SKILL.md — 255 lines: decision framework, question classifier, response rigor
    matrix, statistical philosophy (assumptions-first, effect sizes over p-values)
  • 6 Python scripts: power-analysis, assumption-diagnostics, effect-size-calculator,
    experimental-design, model-comparison, detect-compute (hardware probing)
  • 16 references covering: experimental design, causal inference, Bayesian
    workflow, regression modeling, statistical methodology, PCA workflow, and
    newly added: Docker experiment isolation (301 lines), subagent experiment
    supervision (389 lines), experimental campaign protocol (577 lines, 8-phase),
    PyTorch integration (616 lines), scikit-learn integration (577 lines),
    data science coding workflow (489 lines)
  • 4 test suites — campaign protocol, compute detection, references completeness,
    supervision protocol

  New Skill: epub (epub/)

  Complete EPUB creation, editing, validation, and knowledge extraction skill.
  Built from EPUB 3.3 spec research and real-world testing on a 2.1MB commercial
  Apress EPUB with Apple Books compatibility verification on macOS 26.

  • 11 CLI scripts following cli-builder patterns (--json, --dry-run, lazy deps):
    scaffold, edit (8 subcommands), info, text extraction, knowledge extraction
    (heuristic + LLM auto-detect via env vars), validate, images, batch, convert,
    repair, cover
  • 9 references including new apple-books-compatibility.md — verified rendering
    rules: cover XHTML wrapper requirement, OEBPS/ directory structure, deprecated
    CSS avoidance, spine ordering patterns (3 variants), XHTML namespace handling
  • LLM env var convention — scripts auto-detect EPUB_LLM_URL + EPUB_LLM_KEY
    and enable LLM features automatically; silent fallback to heuristic mode
  • 46/46 tests passing in test_epub_skill.sh

  Data Scientist Expansions

  • Subagent supervision — protocol for spawning worker agents to run experiments
    in parallel with automated health monitoring and result collection
  • Docker experiment isolation — reproducible, containerized data science
    environments with pinned dependency hashing
  • Experimental campaign protocol — 8-phase research workflow from problem
    formulation through publication, with phase-specific templates and completion
    criteria
  • Code integration references — PyTorch, scikit-learn, and general DS coding
    workflow with agent-specific patterns for library usage
  • detect-compute.py — portable hardware probing for ML feasibility (CPU cores,
    RAM, GPU, CUDA toolkit version, disk I/O bandwidth)

  Repository

  • README.md and AGENTS.md updated with both new skills (alphabetical order)
  • All skills follow the Agent Skills open format (agentskills.io)
  • MIT licensed

  Downloads

  • Source code (ZIP)
    2 downloads
  • Source code (TAR.GZ)
    2 downloads