Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Relationship-aware daily gift engine with five-stage creative pipeline — editorial judgment, synthesis, concept generation, visual strategy, and rendering in H5, image, or video
.claude/skills/daily-gift/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | — | — |
| case-08 | ✗→✓ | ▲ Improved | — | — |
| case-18 | ✗→✓ | ▲ Improved | — | — |
| case-03 | ✗→✓ | ▲ Improved | — | — |
| case-17 | ✗→✓ | ▲ Improved | — | — |
A relationship-aware gift engine that decides whether a gift should exist before deciding what it should be. Uses a five-stage creative pipeline to generate personalized daily gifts in H5 (interactive web pages), AI-generated images, or AI-generated videos. The core design principle is "idea before medium" — the creative concept is locked before the output format is chosen.
Published on ClawHub: https://clawhub.ai/jiawei248/daily-gift
Decide whether a gift should exist today, how heavy it should be (skip / nudge / light / standard / heavy), and what content direction to take (reflect, extension, compass, mirror, play, curation, utility, etc.). Format is NOT chosen here.
Extract six content slots from conversation context (today_theme, emotion_peaks, historical_echo, open_loop, lobster_judgment, preference_hint). Form a gift thesis = anchor (which moment deserves the center) + return (what new perspective the agent gives back). If the thesis has no return, it's not a gift — it's a decorated log entry.
Generate 5+ concept candidates using seven thinking angles (metaphor flip, format mashup, impossible action, scale shift, role reversal, time distortion, cultural remix). Cross-pollinate with a library of 73 creative seeds across 8 categories. Run three quality checks: concept quality, concept diversity (8 families), and visual/theme collision detection.
Only after the concept is locked does the system choose the output format (H5, image, or video) based on what best serves the concept.
Choose visual approach, plan assets (pure code, generated background, hybrid), select visual style, and run pre-visualization checks against recent gifts for anti-repetition.
Produce the final artifact. H5 gifts use p5.js/canvas with a quality floor set by built-in templates (300-400 lines of tuned code). Image and video gifts use AI generation APIs. All formats have fallback chains.
User request:
> Decide whether today merits a personalized gift, then create the most fitting concept and output for the recipient.
scripts/ call external APIs (image generation, video generation, asset hosting). Supply API keys only after reviewing which scripts use them.workspace/daily-gift/. No data is sent to external services beyond the configured rendering APIs.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-15 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-21 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-20 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-10 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-11 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-02 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-19 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +55 percentage points is the difference between those two pass rates over the 22 comparable cases.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.