@jairodri/huckleberry-cli 1.0.1 → 1.1.0

package/skills/content-atomizer/references/platforms/twitter.md

@@ -0,0 +1,123 @@
+# Twitter/X
+
+**Algorithm Signals (December 2025):**
+- **Replies** — Highest weight, especially from accounts you engage with
+- **Quote tweets** — 2x engagement value vs plain retweet
+- **Time spent** — On tweet and clicked links
+- **Profile clicks** — Curiosity driven by tweet
+- **Media boost** — Images/videos/GIFs increase visibility scores
+- **Early engagement** — First hours critical for amplification
+
+**2025 Changes:** Following feed now uses **Grok AI** for ranking (based on past interactions and topics) — no longer purely chronological. Users default to For You feed.
+
+**Optimal Specs:**
+
+| Format | Specs | Performance |
+|--------|-------|-------------|
+| Single Tweet | <100 characters optimal | Highest engagement rate |
+| Thread | 8-15 tweets | Best for depth + followers |
+| Quote Tweet | Add value to original | 2x engagement vs retweet |
+| Image Tweet | 1200x675px | 35% more engagement |
+
+## Thread Template
+
+**Tweet 1: Hook Tweet**
+```
+[BOLD CLAIM OR PROMISE]
+
+[What they'll learn in one line]
+
+Thread:
+```
+
+**Tweets 2-X: Content Tweets**
+```
+[NUMBER]. [POINT]
+
+[2-3 sentences of explanation]
+
+[Example or proof if fits]
+```
+
+**Final Tweet: Wrap + CTA**
+```
+TL;DR:
+
+- [Point 1]
+- [Point 2]
+- [Point 3]
+- [Point 4]
+- [Point 5]
+
+If this was useful:
+1. Follow @[handle] for more
+2. RT the first tweet
+
+[Link if relevant]
+```
+
+## Single Tweet Templates
+
+**The Insight Tweet:**
+```
+[Counterintuitive observation about industry/topic]
+
+Most people think [X].
+
+But [Y] is actually true because [Z].
+```
+
+**The List Tweet:**
+```
+[Number] [things] that [outcome]:
+
+- [Item 1]
+- [Item 2]
+- [Item 3]
+- [Item 4]
+- [Item 5]
+
+Which one hits different?
+```
+
+**The Hot Take:**
+```
+Unpopular opinion:
+
+[Contrarian statement]
+
+Here's why: [One-line reasoning]
+```
+
+**The Question Tweet:**
+```
+[Provocative question about industry/topic]?
+
+Genuine question. Reply with your take.
+```
+
+**The Proof Tweet:**
+```
+[Impressive result/stat]
+
+Here's exactly how:
+
+[3-5 bullet points of method]
+```
+
+## Hook Formulas
+
+**Pattern 1: Bold Opener**
+> "[Thing] is dead. Here's what's replacing it:"
+
+**Pattern 2: Numbers + Outcome**
+> "I [did X] for [time period]. Here's what happened:"
+
+**Pattern 3: Controversial Take**
+> "This will piss off [group], but [claim]."
+
+**Pattern 4: Curiosity Gap**
+> "The [industry] secret no one talks about:"
+
+**Pattern 5: Specific Proof**
+> "[Specific result] in [timeframe]. No [common excuse]. Here's the playbook:"

package/skills/content-atomizer/references/platforms/youtube.md

@@ -0,0 +1,56 @@
+# YouTube
+
+**Algorithm Signals (December 2025):**
+- **Click-through rate (CTR)** — How often people click when shown your video
+- **Average view duration & % watched** — Raw watch time AND completion percentage
+- **Session impact** — Does your video keep people on YouTube longer?
+- **Viewer satisfaction** — YouTube now uses surveys + behavior to estimate quality
+- **Negative feedback** — "Not interested," skips, very low retention hurt you
+- **Topical authority** — Channels focused on clear topics get recommended more
+
+**2025 Changes:** AI-driven hyper-personalization. Older evergreen videos get revived when topics trend again. Stronger emphasis on authority, depth, and "entity-rich" content.
+
+**Optimal Specs:**
+
+| Format | Specs | Performance |
+|--------|-------|-------------|
+| Shorts | 10-35 sec (discovery), up to 60 sec | Highest reach, lower depth |
+| Long-form | 8-12 minutes (sweet spot) | Best for monetization + depth |
+| Thumbnail | 1280x720px | CTR target: 4-10% (Home), 6-12% strong |
+
+## Shorts Script Template
+
+```
+[HOOK - 0-2 seconds]
+"[Immediate value promise or pattern interrupt]"
+
+[BODY - 2-50 seconds]
+[Deliver value fast, each point 5-10 seconds max, keep visual movement]
+
+[CTA - 50-60 seconds]
+"Subscribe for more [topic]" OR "Full video on my channel" OR end abruptly (drives rewatch)
+```
+
+## Long-Form Framework (HIVES)
+
+**H - Hook (0-30 seconds)** — Pattern interrupt, credibility, preview. "By the end of this video, you'll know exactly how to [outcome]"
+
+**I - Intro (30 seconds - 1 minute)** — Brief context, who this is for, what you'll cover.
+
+**V - Value (Main content)** — Deliver on the promise. Clear sections with verbal signposting. Examples and proof.
+
+**E - Engagement Prompts (Throughout)** — Every 2-3 minutes: "Let me know in the comments if [question]"
+
+**S - Strong CTA (Final 30 seconds)** — Summarize, clear next action, end screen with subscribe + related video.
+
+## Thumbnail + Title Patterns
+
+**Thumbnail Principles:** 3 elements max (face, text, object). High contrast. Readable at small size. Emotion on face. Curiosity gap.
+
+| Title Pattern | Example |
+|---------------|---------|
+| How I [result] | "How I Built a 6-Figure Newsletter in 8 Months" |
+| [Number] [Things] That [Outcome] | "7 LinkedIn Mistakes Killing Your Reach" |
+| Why [Thing] Doesn't Work | "Why Your Content Strategy Isn't Working" |
+| The [Adjective] [Thing] | "The Boring Marketing Strategy That Actually Works" |
+| I [Did X] For [Time]. Here's What Happened | "I Posted Daily for 90 Days. Here's What Happened" |

package/skills/content-atomizer/references/tweet-patterns.md

@@ -0,0 +1,40 @@
+# 20 Tweet Patterns
+
+A comprehensive taxonomy of tweet types. Use when generating X/Twitter content to ensure structural variety. When generating tweets, randomly select from these patterns rather than defaulting to the same format.
+
+## General Advice / Observation
+1. **Pithy Statement** — Short, impactful advice or observation. "Your network is your net worth."
+2. **Statement + Action** — Advice followed by a call to action. "The best time to start writing was yesterday. The second best time is now. What are you waiting for?"
+3. **Open-Ended Question** — Thought-provoking question inviting reflection. "What's the biggest lesson you learned in your 20s?"
+4. **Direct Quote** — Quote from a famous figure, lending credibility. "The only way to do great work is to love what you do.' - Steve Jobs"
+5. **Quote + Commentary** — Quote followed by your personal interpretation or analysis.
+
+## Tips / Lessons / Insights / Secrets
+6. **Listicle (Bullets)** — Numbered tips in concise bullet points for easy consumption.
+7. **Listicle (Numbered)** — Numbered list with detailed explanations for each point.
+8. **Concept Breakdown** — Specific concept explained in detail with insights, examples, or applications. "The Pareto Principle: 80% of results come from 20% of efforts..."
+9. **Concept + Action** — Concise explanation followed by a specific call to action. "Small, consistent actions create massive results. What habit can you start today?"
+
+## Personal Experience / Story
+10. **Anecdote + Lesson** — Personal experience illustrating a specific lesson, adding authenticity.
+11. **Story Thread** — Longer personal story in multi-tweet thread, building suspense with detailed insights.
+12. **Story + Action** — Personal anecdote followed by a relevant call to action.
+
+## Questions / Polls
+13. **Multiple Choice Poll** — Multiple-choice question with clearly defined options.
+14. **Open-Ended Question** — Question allowing nuanced responses and deeper engagement.
+15. **Question + Context** — Question followed by additional context to encourage thoughtful responses.
+
+## Promotions / Announcements
+16. **Direct Promotion** — Simple announcement with a direct link.
+17. **Promotion + Action** — Announcement followed by a call to action for engagement or lead generation.
+18. **Testimonial + Promotion** — Client testimonial followed by a relevant promotion.
+
+## Humorous / Controversial
+19. **Humorous Statement + Action** — Humorous or controversial statement designed to spark conversation, followed by a CTA.
+20. **Statement + Media** — Humorous or controversial statement paired with a relevant image or video.
+
+## Usage Guidelines
+- When generating a batch of tweets, distribute across at least 4 different pattern categories
+- Match pattern type to content goal (tips for value, stories for connection, humor for engagement)
+- For thread generation, use patterns 6-9 for body tweets, patterns 1-5 for the hook tweet

package/skills/content-repost-manager/SKILL.md

@@ -0,0 +1,245 @@
+---
+name: content-repost-manager
+repo: https://github.com/clawdbot/clawd
+description: >
+  Identify top-performing evergreen content for reposting across platforms.
+  Brand-agnostic — reads analytics data and content taxonomy from any brand's
+  evergreen system folder. Classifies posts as trending (skip) vs evergreen
+  (repost eligible), applies platform-specific cooldowns, and outputs a
+  repost queue. Triggers on "repost", "banger", "top performer", "content
+  recycling", "repost queue", "evergreen content repost", or "which posts
+  should I repost".
+depends-on:
+  - hook-generator
+---
+
+# Content Repost Manager
+
+Brand-agnostic repost intelligence that reads from each brand's evergreen analytics system. Never duplicates existing platform-specific analytics — this is a **lens** on top of existing weekly reports and hook performance data.
+
+## When to Use
+
+- Agent or user asks "which posts should I repost"
+- Weekly content review identifies bangers
+- Cross-platform content recycling (X banger → LinkedIn adaptation)
+- Building a repost queue from analytics data
+
+## Core Principle: Trending vs Evergreen
+
+**Trending posts are NEVER repost candidates.** They reference specific news, product launches, or time-bound events. Reposting stale news destroys credibility.
+
+**Evergreen posts are repost candidates.** They address persistent pain points, share frameworks, or make contrarian observations that stay relevant.
+
+### Classification Rules
+
+| Tag | Category | Repost? |
+|-----|----------|---------|
+| `market_signal` | Trending (usually) | Only if content references a general pattern, not a specific news event |
+| `market_news` | Trending | **Never repost** — news stales in hours |
+| `tool_overload` | Evergreen | Repost eligible |
+| `manual_loops` | Evergreen | Repost eligible |
+| `founder_router` | Evergreen | Repost eligible |
+| `unfinished_workflow` | Evergreen | Repost eligible |
+| `proof_test` | Evergreen | Repost eligible |
+| `context_bloat` | Evergreen | Repost eligible |
+| `template_confusion` | Evergreen | Repost eligible |
+| `expensive_mistake` | Evergreen | Repost eligible |
+| `vertical_law` | Evergreen | Repost eligible |
+| `vertical_real_estate` | Evergreen | Repost eligible |
+| `agent_lifestyle` | Evergreen | Repost eligible |
+| `contrarian` | Evergreen | Repost eligible |
+| `first_person` | Context-dependent | Repost only if not referencing a specific dated event |
+| `pain_pill` | Evergreen | Repost eligible |
+
+When in doubt, read the post text. If it mentions a specific product name, date, or "this week/month" → trending. If it describes a universal pain or framework → evergreen.
+
+## Brand-Agnostic Discovery
+
+This skill works for ANY brand in the evergreen system. The agent must discover the brand's analytics structure by reading its evergreen.md first.
+
+### Step 1: Identify the brand
+
+Read the brand's evergreen.md from:
+```
+businesses/<brand-slug>/evergreen.md
+```
+
+Look for the **Connected Systems** or **File Locations** section to find:
+- Analytics folder path
+- Hook performance data path
+- Content taxonomy definitions
+- Weekly report paths
+
+### Step 2: Read analytics data
+
+Read the latest snapshot from the brand's analytics folder:
+```
+businesses/<brand-slug>/analytics/snapshots/latest-x-metrics.json
+```
+
+Or the latest weekly report:
+```
+businesses/<brand-slug>/analytics/reports/latest-weekly-report.md
+```
+
+If the brand has no analytics folder, check evergreen.md for alternative paths. Some brands store analytics under `content-systems/` or shared directories.
+
+### Step 3: Read content taxonomy
+
+Each brand defines its own pain angles and hook patterns. Read them from:
+- The brand's `content-systems/` folder
+- Or the `weekly-analytics-feedback-loop.md` if it exists for that brand
+- Or the analytics script if it contains classification rules
+
+If no taxonomy exists, extract it from the snapshot data itself (the `pain_angle` and `hook_pattern` fields in JSON snapshots).
+
+## Repost Eligibility Criteria
+
+A post qualifies for repost when ALL of the following are true:
+
+1. **Performance**: Engagement ≥ 2x the brand's median weighted engagement
+2. **Content type**: Classified as evergreen (see table above)
+3. **Cooldown**: Not posted/reposted within the platform's cooldown window
+4. **Not previously reposted**: Check repost history in the queue file
+
+### Platform Cooldowns
+
+| Platform | Cooldown | Rationale |
+|----------|----------|-----------|
+| X/Twitter | 30 days | Fast feed, audience overlap acceptable |
+| LinkedIn | 60 days | Slower feed, different audience subset |
+| Instagram | 45 days | Visual-first, needs fresh image |
+| Medium | 90 days | Long-form, readers expect fresh takes |
+| Threads/Bluesky | 30 days | Fast feed, similar to X |
+
+### Engagement Thresholds
+
+Use the brand's own data. Calculate median weighted engagement from the latest snapshot:
+
+```
+weighted_engagement = likes + (2 × replies) + (2 × reposts)
+```
+
+- **Tier 1 banger**: ≥ 3x median → prioritize for repost + cross-platform adaptation
+- **Tier 2 solid**: ≥ 2x median → queue for repost on same platform
+- **Tier 3 moderate**: ≥ 1.5x median → consider for cross-platform adaptation only
+
+## Cross-Platform Adaptation
+
+When a banger on one platform is adapted for another:
+
+1. **Read the original post** — understand the hook, pain angle, proof, and CTA
+2. **Adapt, never copy** — rephrase for the target platform's conventions:
+   - X → LinkedIn: expand to 2-3 paragraphs, add professional framing
+   - X → Instagram: condense to 1 punchy line + carousel slides
+   - LinkedIn → X: extract the sharpest single insight
+3. **Hook Regeneration** — For Tier 1 bangers, spawn hook-generator subagent for platform-optimized adaptation hooks:
+   ```markdown
+   Load the hook-generator skill and execute the full Phase 1-6 pipeline with this context:
+   - WHO: [Target platform audience from brand evergreen.md]
+   - WHAT/WHY: [Original post's core insight + pain angle]
+   - Knife: [Original post's pain point]
+   - Platform: [Target platform — LinkedIn/Instagram/Threads/etc.]
+   - Brand: [Brand voice from evergreen.md]
+   - Target Score: 80
+   Return ONLY this JSON array: [{"hook":"...","score":<0-100>,"pattern":"...","formula":"..."}]
+   ```
+   Filter: Discard hooks that don't match target platform conventions (e.g., no thread hooks on LinkedIn, no hashtag-heavy hooks on Threads)
+   Fallback: If subagent fails, use manual platform adaptation rules from step 2
+4. **Preserve the winning formula** — keep the same pain angle and hook pattern that performed
+5. **Update the CTA** — match the platform's link conventions
+6. **Tag as `adapted_from: <original_url>`** in the repost queue
+
+## Repost Queue File
+
+Output location (brand-specific):
+```
+businesses/<brand-slug>/analytics/repost-queue.md
+```
+
+### Format
+
+```md
+# Repost Queue — <Brand Name>
+
+> Last updated: YYYY-MM-DD
+> Median engagement: X (from latest weekly report)
+
+## Tier 1 — Bangers (≥ 3x median)
+| Original URL | Platform | Pain | Pattern | Engagement | Repost Eligible On | Adapted For | Status |
+|---|---|---|---|---:|---|---|---|
+| https://x.com/... | X | tool_overload | contrarian | 22 | 2026-06-15 | LinkedIn | queued |
+
+## Tier 2 — Solid (≥ 2x median)
+| Original URL | Platform | Pain | Pattern | Engagement | Repost Eligible On | Adapted For | Status |
+|---|---|---|---|---:|---|---|---|
+
+## Tier 3 — Cross-Platform Candidates (≥ 1.5x median)
+| Original URL | Platform | Pain | Pattern | Engagement | Adapt For | Status |
+|---|---|---|---|---:|---|---|
+
+## Repost History
+| Date | Original URL | Platform | Action | Result |
+|---|---|---|---|---|
+```
+
+### Status Values
+
+- `queued` — waiting for cooldown to expire
+- `ready` — cooldown passed, ready to post
+- `adapted` — rewritten for different platform
+- `posted` — live, track results
+- `killed` — decided not to repost after all
+- `skip` — content staled or became irrelevant
+
+## Integration with Existing Systems
+
+This skill is a **consumer** of analytics data, not a producer. It reads from:
+
+| Data Source | Owner | What We Read |
+|-------------|-------|-------------|
+| Weekly analytics script | Hermes / analytics cron | JSON snapshots, weekly reports |
+| Hook performance registry | Hook intelligence skill | Winning hook patterns |
+| Content taxonomy | Brand evergreen.md | Pain angles, hook patterns, CTA types |
+| Posting history | Platform posting queues | What was already posted/reposted |
+
+It writes to:
+| Output | What |
+|--------|------|
+| `analytics/repost-queue.md` | Repost candidates + history |
+| Evergreen.md systems table | Entry pointing to this skill + queue file |
+
+## Workflow
+
+### Manual (on demand)
+
+When user asks "what should I repost" or "which posts are bangers":
+
+1. Read the brand's evergreen.md → find analytics paths
+2. Read latest snapshot or report
+3. Calculate median engagement
+4. Filter for evergreen content types only
+5. Apply cooldown check (compare post date vs today)
+6. Output the repost queue
+
+### Automated (cron extension)
+
+Add a repost scan step to the existing weekly analytics cron:
+
+```bash
+# After the existing analytics script runs, scan for repost candidates
+python3 scripts/content-repost-scan.py --business <brand-slug>
+```
+
+This script reads the snapshot output by the analytics script and appends to the repost queue.
+
+## Anti-Patterns
+
+| Mistake | Why Wrong | Fix |
+|---------|-----------|-----|
+| Reposting market news | Stale in hours, destroys credibility | Classify first, never repost trending |
+| Same text on same platform | Audience remembers, looks lazy | Always adapt or add new framing |
+| 30-day cooldown on news | Irrelevant — news is dead by then | No cooldown needed, just don't repost |
+| Ignoring cross-platform opportunity | X banger may kill it on LinkedIn | Always check Tier 1 for adaptation |
+| Duplicating analytics logic | Hermes already runs weekly analytics | Read data, don't regenerate |
+| Brand-specific hardcoded paths | Breaks for other brands | Always discover from evergreen.md |

package/skills/content-repost-manager/scripts/content-repost-scan.py

@@ -0,0 +1,133 @@
+#!/usr/bin/env python3
+"""Content Repost Scanner: Reads analytics snapshots and identifies repost candidates."""
+import json
+import sys
+import os
+from datetime import datetime, timedelta
+
+def load_snapshot(path):
+    with open(path) as f:
+        return json.load(f)
+
+def calc_weighted_engagement(post):
+    return post.get('likes', 0) + 2 * post.get('replies', 0) + 2 * post.get('reposts', 0)
+
+def classify_evergreen(post):
+    tag = post.get('pain_angle', post.get('tag', '')).lower()
+    trending_tags = {'market_signal', 'market_news'}
+    evergreen_tags = {
+        'tool_overload', 'manual_loops', 'founder_router', 'unfinished_workflow',
+        'proof_test', 'context_bloat', 'template_confusion', 'expensive_mistake',
+        'vertical_law', 'vertical_real_estate', 'agent_lifestyle', 'contrarian',
+        'pain_pill', 'first_person',
+    }
+    if tag in trending_tags:
+        return False
+    if tag in evergreen_tags:
+        return True
+    return 'evergreen' in tag
+
+def main():
+    import argparse
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--business', required=True, help='Brand slug')
+    parser.add_argument('--snapshot', help='Path to analytics snapshot JSON')
+    parser.add_argument('--output', help='Output queue file path')
+    args = parser.parse_args()
+
+    business = args.business
+    base = f'businesses/{business}'
+
+    snapshot_path = args.snapshot or f'{base}/analytics/snapshots/latest-x-metrics.json'
+    if not os.path.exists(snapshot_path):
+        print(f'No snapshot found at {snapshot_path}')
+        sys.exit(0)
+
+    snapshot = load_snapshot(snapshot_path)
+    posts = snapshot if isinstance(snapshot, list) else snapshot.get('posts', [])
+
+    if not posts:
+        print('No posts in snapshot')
+        return
+
+    engagements = [calc_weighted_engagement(p) for p in posts]
+    median = sorted(engagements)[len(engagements) // 2] if engagements else 0
+
+    candidates = []
+    for post in posts:
+        ew = calc_weighted_engagement(post)
+        ratio = ew / median if median > 0 else 0
+        if not classify_evergreen(post):
+            continue
+        post_date = post.get('date', post.get('created_at', ''))
+        if post_date:
+            try:
+                pd = datetime.fromisoformat(post_date.replace('Z', '+00:00'))
+                days_ago = (datetime.now().astimezone() - pd).days
+            except:
+                days_ago = 999
+        else:
+            days_ago = 999
+
+        candidates.append({
+            'url': post.get('url', ''),
+            'platform': post.get('platform', 'X'),
+            'pain': post.get('pain_angle', post.get('tag', 'unknown')),
+            'pattern': post.get('hook_pattern', 'unknown'),
+            'engagement': ew,
+            'ratio': round(ratio, 1),
+            'age_days': days_ago,
+            'eligible_after': days_ago,
+        })
+
+    candidates.sort(key=lambda c: c['engagement'], reverse=True)
+
+    tier1 = [c for c in candidates if c['ratio'] >= 3]
+    tier2 = [c for c in candidates if c['ratio'] >= 2 and c not in tier1]
+    tier3 = [c for c in candidates if c['ratio'] >= 1.5 and c not in tier1 and c not in tier2]
+
+    output_lines = [f'# Repost Queue — {business}', '']
+    output_lines.append(f'> Last updated: {datetime.now().strftime("%Y-%m-%d")}')
+    output_lines.append(f'> Median engagement: {median}')
+    output_lines.append('')
+
+    if tier1:
+        output_lines.append('## Tier 1 — Bangers (≥ 3x median)')
+        output_lines.append('| URL | Platform | Pain | Pattern | Engagement | Ratio | Age (days) | Eligible | Status |')
+        output_lines.append('|---|---|---|---:|---:|---:|---|---|')
+        for c in tier1:
+            eligible_on = (datetime.now() + timedelta(days=max(0, 30 - c['age_days']))).strftime('%Y-%m-%d')
+            output_lines.append(f"| {c['url']} | {c['platform']} | {c['pain']} | {c['pattern']} | {c['engagement']} | {c['ratio']}x | {c['age_days']} | {eligible_on} | queued |")
+        output_lines.append('')
+
+    if tier2:
+        output_lines.append('## Tier 2 — Solid (≥ 2x median)')
+        output_lines.append('| URL | Platform | Pain | Pattern | Engagement | Ratio | Age (days) | Eligible | Status |')
+        output_lines.append('|---|---|---|---:|---:|---:|---|---|')
+        for c in tier2:
+            output_lines.append(f"| {c['url']} | {c['platform']} | {c['pain']} | {c['pattern']} | {c['engagement']} | {c['ratio']}x | {c['age_days']} | — | queued |")
+        output_lines.append('')
+
+    if tier3:
+        output_lines.append('## Tier 3 — Cross-Platform Candidates (≥ 1.5x median)')
+        output_lines.append('| URL | Platform | Pain | Pattern | Engagement | Ratio | Adapt For | Status |')
+        output_lines.append('|---|---|---|---:|---:|---|---|')
+        for c in tier3:
+            alt_platform = 'LinkedIn' if c['platform'] == 'X' else 'X'
+            output_lines.append(f"| {c['url']} | {c['platform']} | {c['pain']} | {c['pattern']} | {c['engagement']} | {c['ratio']}x | {alt_platform} | candidate |")
+        output_lines.append('')
+
+    output = '\n'.join(output_lines)
+
+    output_path = args.output or f'{base}/analytics/repost-queue.md'
+    os.makedirs(os.path.dirname(output_path), exist_ok=True)
+    with open(output_path, 'w') as f:
+        f.write(output)
+
+    print(f'Repost queue written to {output_path}')
+    print(f'Tier 1 (bangers): {len(tier1)}')
+    print(f'Tier 2 (solid): {len(tier2)}')
+    print(f'Tier 3 (cross-platform): {len(tier3)}')
+
+if __name__ == '__main__':
+    main()

package/skills/context-recovery/.clawdhub/origin.json

@@ -0,0 +1,7 @@
+{
+  "version": 1,
+  "registry": "https://clawhub.ai",
+  "slug": "context-recovery",
+  "installedVersion": "1.2.2",
+  "installedAt": 1772816764944
+}