fix(skills): put trigger phrasing in every skill description

[TMHSDigital]

Closes #357. Closes #396.

Problem

Claude Code and Cursor decide whether to load a skill from its frontmatter description alone.

• All 16 skills kept their triggers in a body ## Trigger section that the model can't see at selection time.
• 0 of 16 descriptions had a "Use when…" clause.
• The documented "under 200 chars" limit (AGENTS.md, CONTRIBUTING.md) was already broken by 12 skills, was never enforced, and would have blocked the fix.

Change

• Every description now has two parts:
  • what the skill covers (unchanged in spirit; this is what the landing cards show);
  • then "Use when …" with the API names and literal error strings that should load it. Examples: 'No active UV layer found', poll() failed, context is incorrect, AttributeError on sequence_editor.sequences, 'Node type ... undefined', keys that never play back.
• Descriptions are written as YAML double-quoted strings, so colons in error text are safe.
• Two descriptions also fix stale content:
  • slotted-actions-animation no longer recommends bare action.fcurves on 4.5 (snippets: Slotted Actions bridge makes keys that never play back on 4.5 LTS (no slot assigned on the legacy path) #385);
  • engine-export-presets matches snippets: Godot glTF preset exports Z-up, but glTF and Godot are both Y-up #346 (Y-up, meter glTF for every engine). This overlaps fix(snippets): export Y-up, meter glTF for Godot and Unreal too #417's description line; whichever merges second gets rebased.
• tests/check_frontmatter.py: each skill description must be ≤ 1024 characters (the Agent Skills format cap; Claude Code's own cap is 1536 for description + when_to_use) and must contain "Use when".
• AGENTS.md / CONTRIBUTING.md: the example frontmatter shows the new shape instead of "under 200 chars".

Evidence (local; no Blender needed)

• python tests/check_frontmatter.py: skill and rule frontmatter valid. All 16 descriptions yaml.safe_load cleanly. Longest is 501 chars.
• Falsified: with "Use when" replaced by "Handy when" in ui-panels, the check prints ::error::skills/ui-panels/SKILL.md: description has no 'Use when ...' trigger clause and exits 1. Restored, it passes.
• Landing page: build_site._summarize() (340-char, whole-sentence cut) shows only the first, "what it covers" sentence(s) for every skill (128–293 chars). The trigger clause doesn't clutter the cards.

Not measured: the change in skill-selection rate. That is what the eval harness in #373 would quantify.

🤖 Generated with Claude Code

[TMHSDigital]

CI evidence before merge: all checks pass ( 13 pass ). Blender Smoke on PR head: Blender 5.2.2 LTS / Blender 4.5.14 LTS. Summaries (happy path; falsifiers) per leg: 141 passed, 3 skipped, 0 failed;136 passed, 6 skipped, 0 failed;143 passed, 1 skipped, 0 failed;142 passed, 0 skipped, 0 failed;

[TMHSDigital]

CI evidence before merge: all checks pass ( 13 pass ). Blender Smoke on PR head: Blender 5.2.2 LTS / Blender 4.5.14 LTS. Summaries (happy path; falsifiers) per leg: 143 passed, 1 skipped, 0 failed;142 passed, 0 skipped, 0 failed;141 passed, 3 skipped, 0 failed;136 passed, 6 skipped, 0 failed;