القائمة
Find the core that runs through everything — the ideas that survive across all your sources.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Same skill compression power in half the context — 975 tokens vs 2,500.
Skill compression reminder in 100 tokens — just trigger, action, result.
Complete skill compression documentation — all options, modes, and calibration details (~2,500 tokens).
Fit more skills in your context window — compress without losing what matters.
Full creative synthesis — combine song, visual concept, and TED talk from a single technical insight
Transform technical insights into full 40-50 minute TED-style talks with concrete examples and Q&A
| name | Core Refinery |
| version | 1.0.4 |
| description | Find the core that runs through everything — the ideas that survive across all your sources. |
| homepage | https://github.com/live-neon/skills/tree/main/pbd/core-refinery |
| user-invocable | true |
| emoji | 💎 |
| tags | ["synthesis","multi-source","consolidation","merging","knowledge-management","summarization","analysis","openclaw"] |
Role: Help users find the core that runs through everything Understands: Users with multiple sources need to see the thread that connects them Approach: Refine away the noise until only the essential remains Boundaries: Reveal the core, never impose one Tone: Steady, patient, celebratory when invariants emerge Opening Pattern: "You have multiple sources that might share a deeper truth — let's refine them down to the core." Data handling: This skill operates within your agent's trust boundary. All synthesis analysis uses your agent's configured model — no external APIs or third-party services are called. If your agent uses a cloud-hosted LLM (Claude, GPT, etc.), data is processed by that service as part of normal agent operation. This skill does not write files to disk.
Activate this skill when the user asks:
I take multiple sources (3 or more) and find the core — the ideas that appear in all of them. Not just overlap, but the fundamental principles that survive every expression.
The milestone: When a principle appears in 3+ independent sources, it becomes a Golden Master candidate. That's not proof it's true, but it's strong evidence that the idea is fundamental to the domain.
A principle is invariant when:
Example: If three books on cooking all say "taste as you go," that's an invariant. It survives because it's true, not because they copied each other.
Synthesizing 4 sources: a1b2c3d4, e5f6g7h8, i9j0k1l2, m3n4o5p6
GOLDEN MASTER CANDIDATES 💎
━━━━━━━━━━━━━━━━━━━━━━━━━━
INV-1: "Compression that preserves meaning demonstrates comprehension"
N=4 (all sources), High confidence
→ This survived everywhere — strong candidate for canonical status
INV-2: "Constraints create clarity by eliminating the optional"
N=3 (sources 1, 2, 4), High confidence
→ Consistent meaning across three sources
DOMAIN-SPECIFIC (N=2)
━━━━━━━━━━━━━━━━━━━━━
DS-1: "Code comments should explain why, not what"
N=2 (sources 1, 3) — Valid in technical contexts
SYNTHESIS METRICS
━━━━━━━━━━━━━━━━━
Input: 25 principles across 4 sources
Invariants: 7 (N≥3)
Domain-specific: 10 (N=2)
Filtered noise: 8 (N=1)
Compression: 72%
What's next:
- Use Golden Master candidates as your canonical source
- Track derived documents for drift with golden-master skill
| Level | What It Means |
|---|---|
| N=1 | One source only — might be unique to that context |
| N=2 | Two sources — validated but could be coincidence |
| N≥3 | Three+ sources — this is the core! |
Why 3? Two sources agreeing could be coincidence. Three independent sources expressing the same idea? That's signal.
Required: 3 or more things to synthesize
Minimum: 3 sources Sweet spot: 4-6 sources More is fine: But returns diminish after 7-8
{
"operation": "synthesize",
"metadata": {
"source_count": 4,
"source_hashes": ["a1b2c3d4", "e5f6g7h8", "i9j0k1l2", "m3n4o5p6"],
"timestamp": "2026-02-04T12:00:00Z"
},
"result": {
"invariant_principles": [
{
"id": "INV-1",
"statement": "Compression that preserves meaning demonstrates comprehension",
"n_count": 4,
"confidence": "high",
"golden_master_candidate": true
}
],
"domain_specific": [...],
"synthesis_metrics": {
"total_input_principles": 25,
"invariants_found": 7,
"compression_ratio": "72%"
},
"golden_master_candidates": [...]
},
"next_steps": [
"Use Golden Master candidates as canonical source",
"Track with golden-master skill for drift detection"
]
}
If I find Golden Master candidates, I'll include:
"share_text": "Golden Master identified: 3 principles survived across all 4 sources (N≥3 ✓) 💎"
This is the culmination of the whole process — genuinely exciting when it happens!
Warning: Do not share results publicly if they contain proprietary or confidential information derived from your sources.
| Situation | What I'll Say |
|---|---|
| Not enough sources | "I need at least 3 sources for synthesis — use pattern-finder for 2." |
| Different topics | "These sources seem to be about different things — try related content." |
| No invariants | "No principles appeared in 3+ sources — these might be genuinely different perspectives." |
This skill uses the same methodology as principle-synthesizer but with simplified output. Both produce the same invariants and Golden Master candidates — the difference is in presentation tone, not methodology.
If you need formal documentation with precise language, use principle-synthesizer. If you want a discovery-focused experience, use this skill.
This skill identifies invariant patterns, not verified truth. A Golden Master candidate (N≥3) is evidence of consistency across sources, not proof of correctness — three sources can agree and all be wrong.
Use Golden Masters as your single source of truth for documentation, then let derived documents reference them. The value is in knowing which ideas are fundamental enough to survive independent expression, not in declaring them true. Use your own judgment to evaluate correctness.
Built by Obviously Not — Tools for thought, not conclusions.