Value
The 64 runnable examples are the repo's strongest asset, but the skills barely point to them, so neither agents nor humans find the proof for a given skill.
Evidence
- 10 of 16 skills link no
examples/ path.
slotted-actions-animation never mentions turntable, even though turntable exists to show the channelbag bridge.
headless-batch-scripting never mentions temp-override-join.
geometry-nodes-python names 2 of its 7 GN examples.
- An agent has to join three overlapping manifests:
examples/gallery.json (56 entries), tests/smoke/catalog.json (142 rows, with names like vse-cut-list-pixels ≠ vse-cut-list), and .cursor-plugin/plugin.json.
- Catalog
min_version exists on only a few rows, and gallery detail pages show no supported-version range.
Suggested approach
- Add a
skills: [...] field per example.
- Generate
examples/index.json with name, path, skills, min_version, render/check-only, try-it command, every falsifier flag with its expected exit, and the exit-code table.
- Generate a "Runnable examples" section into each SKILL.md from the index, and fail validation when a skill has none.
- Render 4.5 / 5.1 / 5.2 badges and a "falsify it" copy line on gallery detail pages.
Done when
Every skill links its examples automatically, and one JSON file describes all examples.
Value
The 64 runnable examples are the repo's strongest asset, but the skills barely point to them, so neither agents nor humans find the proof for a given skill.
Evidence
examples/path.slotted-actions-animationnever mentionsturntable, even though turntable exists to show the channelbag bridge.headless-batch-scriptingnever mentionstemp-override-join.geometry-nodes-pythonnames 2 of its 7 GN examples.examples/gallery.json(56 entries),tests/smoke/catalog.json(142 rows, with names likevse-cut-list-pixels≠vse-cut-list), and.cursor-plugin/plugin.json.min_versionexists on only a few rows, and gallery detail pages show no supported-version range.Suggested approach
skills: [...]field per example.examples/index.jsonwith name, path, skills,min_version, render/check-only, try-it command, every falsifier flag with its expected exit, and the exit-code table.Done when
Every skill links its examples automatically, and one JSON file describes all examples.