Die Oberfläche ist lokalisiert; Skill-Beschreibungen und -Dokumente werden per KI automatisch in Ihre Sprache übersetzt. Code, Befehle, Pfade und Fachbegriffe bleiben im Original, maßgeblich ist das Dokument des Autors.
Skills entdecken
Daten und Forschung · SKILL.md

repo-maintenance

repo-maintenance

Maintainer workflow router for editing PyOD source, tests, docs, packaging metadata, CLI entry points, and the packaged od-expert skill safely.

Independent GitHub snapshot · Automated classification · Original documentation governs

Skill-Dokumentation

GitHub

name: repo-maintenance description: "Maintainer workflow router for editing PyOD source, tests, docs, packaging metadata, CLI entry points, and the packaged od-expert skill safely." disable-model-invocation: true metadata: disco-role: operating license: BSD 2-Clause

PyOD Repo Maintenance

Use this sub-skill when the user is contributing to PyOD itself: editing source layout, package metadata, docs examples, CLI wiring, tests, or the packaged od-expert agent skill. This route is for maintainer work in a PyOD checkout, not for running anomaly detection as an end user.

Route here for

  • Locating PyOD source roots, package data, optional extras, or entry points.
  • Choosing focused tests after edits to detectors, utilities, CLI, MCP import safety, persistence, thresholding, documentation, or packaged skills.
  • Updating the shipped od-expert skill or KB-derived skill sections.
  • Maintaining scripts/regen_skill.py, pyod/skills/__init__.py, or the package-data entries that make skills installable.
  • Updating docs examples, especially the agentic demo figure workflow.
  • Checking repo surface health without mutating files.

Route elsewhere

  • End-user detector fitting, scoring, and tabular APIs: classic-detectors.
  • ADEngine, pyod info, pyod install skill, or MCP usage as a user workflow: automated-lifecycle.
  • Time-series, graph, text/image/audio, and optional backend use: specialized-modalities.
  • Persistence, thresholding, score combination, and operational model artifacts: model-operations.
  • Release publishing, credentials, PyPI uploads, or destructive release actions: do not proceed unless an authorized PyOD maintainer explicitly approves the exact command and target.

First-pass maintainer workflow

  1. Identify the edited area and read the closest bundled reference below.

  2. Run the safe surface checker before proposing commands that assume files or entry points exist:

    python path/to/this/sub-skill/scripts/repo_surface_check.py --repo-root <pyod-checkout>
    

    Add --import-check only when importing PyOD from the checkout is intended.

  3. Pick a focused test slice from references/testing-guide.md; do not default to all optional-extra tests unless the edit changed those extras.

  4. For od-expert content, run the generator check before and after editing KB-derived material. Only hand-written sections should be edited directly.

  5. If a workflow would publish, upload, delete, or use credentials, stop and ask for maintainer approval instead of inferring authorization.

Bundled references and scripts

  • references/repo-layout.md — read when locating source roots, package-data ownership, optional dependencies, or CLI entry points.
  • references/testing-guide.md — read when selecting non-destructive, focused pytest/CLI/generator checks by changed area.
  • references/skill-maintenance.md — read before editing the packaged od-expert skill, KB-derived sections, installer maps, or packaged skill tests.
  • references/source-script-inventory.md — read when deciding whether a repo script/example should be copied, adapted, wrapped, left reference-only, or excluded.
  • references/troubleshooting.md — read when a maintainer check fails, an optional dependency is missing, packaged skill drift is reported, or a docs rendering helper fails.
  • scripts/repo_surface_check.py — safe diagnostic helper; reports key files, packaging metadata, relevant tests, and optional command/import availability without modifying the checkout.

Guardrails

  • Keep commands relative to the current PyOD checkout. Do not use private environment paths or machine-specific directories in public instructions.
  • Prefer python -m pytest ... and python -m pyod.cli ... so the selected Python environment is explicit.
  • Treat optional extras (torch, graph, embedding, openai, huggingface, audio, mcp, suod, xgboost, combo, pythresh) as optional unless the changed area requires them.
  • Do not edit generated KB-derived Markdown blocks by hand; update the knowledge source/generator and regenerate.
  • Do not bundle or duplicate the repo's release commands. Publishing requires a maintainer-owned release procedure and explicit approval.
Gespeicherte Skills