An Agent Skill that builds genuine expertise in any topic — and compounds it across sessions.
Most "learn a domain" skills summarize a field. /deepen doesn't. Point it at a topic and it acquires defensible expertise: it weights sources by checkable track record (not fame), takes falsifiable stances on the open controversies, attacks its own conclusions before committing them, and writes the result as an evergreen knowledge base that gets deeper every time you run it.
/deepen PPC
/deepen sleep apnea
/deepen pickleball paddle physics
Works anywhere the Agent Skills open standard is supported — Claude Code, Claude.ai, the Claude API, and compatible runtimes.
| Typical "domain expert" skill | /deepen |
|---|---|
| Ranks sources by authority (official docs → blogs) | Ranks by checkable track record — a famous source with bad calls loses to an independent one with good ones |
| Reports "both sides have merit" | Takes a stance with calibrated confidence + the explicit evidence that would flip it |
| States conclusions as fact | Cite-or-flag: every claim cites a source or is tagged [inference]/[convention] — no confident hallucination |
| Simulates a fictional "expert persona" | Synthesizes the real published record — never fabricates experts, credentials, or quotes |
| One-shot report | Compounds: a living knowledge base that deepens on every re-run, with a changelog and supersede rules |
A competent literature summary tells you what people say. Expertise tells you what's true, what's contested, what you judge correct, and what would change your mind.
The workflow is built on a verified research synthesis:
- Cooke's Classical Model / structured expert judgment (Colson & Cooke 2018) — weight experts by demonstrated accuracy; across 33 studies, <1/3 of credentialed experts were statistically accurate.
- Systematic overconfidence (Morgan 2014, PNAS) — experts' stated uncertainty is consistently too narrow, so confidence intervals get widened on purpose.
- CFAR Double Crux — a procedure for finding the load-bearing belief under a disagreement and taking a falsifiable position.
- Andy Matuschak's evergreen notes — atomic, concept-oriented, densely-linked notes designed to accrete.
- Anthropic's orchestrator-worker pattern — decompose, fan out parallel researchers, synthesize (~+90% over single-agent on breadth-first research).
Where a step rests on practitioner convention rather than verified evidence (e.g. ICE/RICE tactic ranking), the skill says so — and so does its output.
A folder per topic under your knowledge-base root (./knowledge/<topic>/ by default, or set DEEPEN_KB):
knowledge/ppc/
_map.md # index + depth-calibration dashboard (what's solid vs. thin)
principles.md # durable first-principles, each with a "common misconceptions" line
experts.md # the expert map, weighted by track-record proxy ladder
disagreements.md # live debates: crux, my stance, confidence, falsifier, survived-refutation
tactics.md # ranked A+ playbook, tuned to YOUR context
changelog.md # one line per run — accretion stays auditable
Every run ends with a tight brief: what I now know (with confidence), where I took a stance (and what would flip it), the ranked tactics, and the thin spots worth a deeper pass.
examples/deliberate-practice/ is a genuine /deepen run, committed verbatim — real citations, falsifiable stances on a live expert disagreement (Ericsson vs. the Macnamara/Hambrick meta-analyses), and one debate deliberately held at LOW confidence and flagged as thin. Start with disagreements.md.
ClawHub (OpenClaw and compatible agents):
clawhub install deepenGit / Claude Code — clone or copy the skill folder into your skills directory:
git clone https://github.com/Brianhkerr/deepen-skill.git
cp -r deepen-skill/skills/deepen ~/.claude/skills/deepenOr just copy skills/deepen/SKILL.md into your agent's skills directory (e.g. ~/.claude/skills/deepen/ for Claude Code).
/deepen <topic>
- Light pass (default): a fast, mostly-from-knowledge map — good for low-stakes topics or a first sketch.
- Deep pass: full parallel-research fan-out with the adversarial self-falsification gate — for high-value topics you want to actually master.
Re-run it on the same topic any time; it deepens the existing knowledge base instead of starting over, targeting whatever's still below your target depth.
- Track record over credentials. Authority is a weak prior, never the ranking.
- Take falsifiable stances. A wrong-but-falsifiable position is recoverable; a mushy non-answer isn't.
- Red-team yourself. Stances must survive an adversarial refutation, a teach-back, and a prediction test before they're "earned."
- Honesty about evidence strength. Cite-or-flag everything; never fabricate.
- Compound, don't restart. The knowledge base is the product.
MIT-0 (MIT No Attribution) — see LICENSE. Free to use, modify, and redistribute, no attribution required.