Inspiration Library

Action:

1. Parse the current conversation context to extract:

   • Title — descriptive name for the entry
   • Type — auto-infer from context: transcript → Transcript, creator analysis → Creator Pattern, hook → Hook Example, research data → Research Finding, script format → Script Format, carousel/slides → Carousel Example
   • Platform — infer from context or ask if unclear
   • Hook Framework — match to one of: Curiosity Gap, Proof-First, Pain Point, Contrarian, This Cost Me Thousands, Unexpected Confession, Question Hook, Pattern Interrupt, Before vs After, Test / Experiment
   • Performance Tier — Outlier (3x+ median, 100K+ views), Strong (above-avg with metrics), Reference (useful, no metrics), Untested (no data)
   • Source — Creator Analysis, Research, Transcript, My Content, Manual
   • Source URL — if a URL exists in context
   • Creator Handle — @handle if present
   • Topic Tags — infer from content, use existing tags when possible: AI Tools, Business Automation, Productivity, Tutorial, Social Proof. New tags are fine.
   • Engagement Metric — any numbers: views, likes, revenue, growth
   • Hook Text — the actual hook/template if applicable
   • Notes — any additional context worth preserving. For Carousel Example entries, Notes MUST include: slide count, design style (gradient/flat/textured/photo), color palette, text density (heavy/medium/light), photo vs graphic, typography style, and any distinctive layout patterns

2. Aim for under 3 questions before creating. Auto-infer as much as possible. Only ask when truly ambiguous (e.g., Type could be multiple things, or Performance Tier is unclear).

3. Show the user what you're about to save:

   Saving to Inspiration Library:

   • Title: {title}
   • Type: {type} | Tier: {tier}
   • Hook Text: "{hook_text}"
   • Tags: {tags}

   Good to save?

4. Create the Notion page using notion-create-pages with data_source_id from config:

   • Set date:Date Added:start to today's date (ISO-8601)
   • If there's substantial context (full transcript, detailed breakdown), put it in the page body
   • Properties only get the compact data

5. Confirm: "Saved: {Title} ({Type}, {Tier})"

Browse / Query

Trigger: "show my library", "browse", "inspiration on [topic]", "what hooks do I have for [framework]", "show [type] entries", "library"

Action:

1. Use notion-search with data_source_url set to collection://{inspiration_data_source_id} to find matching entries
2. For topic queries, search by the topic keyword
3. For type/framework/tier queries, search and filter the results

Output format — compact list:

INSPIRATION LIBRARY ({filter context}):

[1] "{Hook Text or Title}" — @{creator} | {Type} | {Tier} | {Engagement Metric}
[2] "{Hook Text or Title}" — @{creator} | {Type} | {Tier} | {Engagement Metric}
...

{N} entries found. Say "show me #N" for full details.

If user asks for full details on a specific entry, fetch the page and show all properties + page body content.

Manual Add

Trigger: "add [URL] to library", "save this link", URL + "inspiration" or "library"

Action:

1. If the URL is a video (TikTok, Instagram, YouTube, X, Facebook):
   • Offer: "Want me to pull the transcript first? I can extract the hook and key patterns."
   • If yes, use the /transcript skill flow (read ~/.claude/skills/transcript/skill.md), then save with Type=Transcript, Source=Transcript
   • Auto-extract: Hook Text = first spoken line, Creator Handle from URL
2. If non-video URL: save as Type=Research Finding, Source=Manual
3. Follow the same save flow as above

Score Research

Trigger: "score this research", "analyze these findings", "which ones should I save", "best findings", "curate"

Action: When research output exists in the conversation (from /last30days or any research), score each finding on 5 dimensions:

Rate each finding 0-5 stars total. Show ranked output:

RESEARCH SCORING:

[1] ★★★★★ "{quote}" — {engagement}, {source}
    → {why it scored high}. SAVE AS: {entry type}
[2] ★★★★☆ "{quote}" — {engagement}, {source}
    → {assessment}. SAVE AS: {entry type}
[3] ★★★☆☆ "{quote}" — {engagement}, {source}
    → {assessment}. Not library-worthy yet.

PATTERNS: "{theme}" cluster ({N} findings)

RECOMMENDED SAVES: {N} entries. Save them?

If user confirms, save qualifying entries using the standard Save Entry flow.

Library Health Check

Trigger: "library health", "how fresh is my library", "what's stale", "audit my library", "coverage gaps"

Action:

1. Query ALL entries from the Inspiration Library (up to 50 results)
2. Analyze and report:

LIBRARY HEALTH:

Freshness: {N}/{total} entries added in last 30 days
Coverage gaps: Missing entries for {topics with 0 entries}
Tier balance: {N} Outlier, {N} Strong, {N} Reference, {N} Untested
Framework gaps: {frameworks with <2 examples}
Stale (never used, 60+ days old): {N} entries

ACTIONS:
- {specific recommendation}
- {specific recommendation}

Analysis dimensions:

• Age — entries by Date Added, flag anything 90+ days with no Last Used
• Coverage — Topic Tags distribution vs common content topics
• Tiers — warn if >50% are Reference/Untested
• Frameworks — flag any of the 10 hook frameworks with <2 examples
• Stale — entries with no Last Used date AND Date Added > 60 days ago

Stats

Trigger: "how's my library", "library stats", "stats", "how many entries"

Action:

1. Query all entries from the inspiration library
2. Compile and show:

INSPIRATION LIBRARY STATS:

Total entries: {N}

By Type:
- Hook Example: {n}
- Creator Pattern: {n}
- Script Format: {n}
- Transcript: {n}
- Research Finding: {n}
- Carousel Example: {n}
- Other: {n}

By Performance Tier:
- Outlier: {n}
- Strong: {n}
- Reference: {n}
- Untested: {n}

Top Hook Frameworks:
- {framework}: {n} entries
- {framework}: {n} entries
...

Most-used Topic Tags:
- {tag}: {n} entries
...

Last added: {date} — "{title}"
Last used in scripts: {date} — "{title}"

Query Modes (for other skills)

When other skills query this library during content creation, they should use these modes:

Mode 1: Property-Only (default during script creation)

Returns: Title, Type, Hook Framework, Engagement Metric, Hook Text. For carousel queries, also return Notes field (contains style data needed for remix generation). Max 5 results. ~250 tokens total.

Compact injection format:

INSPIRATION LIBRARY ({filter context}):
[1] "{Hook Text}" — @{creator} | {Tier} | {Engagement Metric}
[2] "{Hook Text}" — @{creator} | {Tier} | {Engagement Metric}

Mode 2: Summary (research phase)

Up to 10 entries, properties only. For supplementing research output.

Mode 3: Full Page (on-demand only)

Full page body content. Only when user explicitly asks "show me the transcript for X" or "full details on X". Never auto-loaded during script creation.

Conversation Style

• Quick and direct. "Saved: Harper Carroll series pattern (Creator Pattern, Outlier)" — not a paragraph about what you did.
• Auto-infer aggressively. If context makes Type, Tier, and Tags obvious, don't ask. Just show what you're saving and let them correct.
• Max 2 clarifying questions. If you need more than that, save what you can and note the gaps.
• Compact output. Library browse results should be scannable, not walls of text.

Opening (No Arguments)

If invoked without arguments:

What do you need?

I can:

• Save — "save this as inspiration" (uses current context)
• Browse — "show hooks for Proof-First" or "inspiration on AI tools"
• Add URL — "add [URL] to library" (optionally pulls transcript)
• Score research — "which findings should I save?"
• Health check — "how fresh is my library?"
• Stats — "how's my library"

Or just tell me what you're looking at.