Claude Code skills I actually run, shared as-is. Each folder is one skill: a SKILL.md you can
drop into ~/.claude/skills/<name>/, plus a README.md explaining the thinking behind it.
These are patterns extracted from real daily use — the anecdotes in each write-up happened.
| Skill | One line | Language |
|---|---|---|
dialectic |
Harden a plan/design/decision with independent AI reviewers that don't know what you want — an unprimed thesis→antithesis→synthesis (정반합) loop. | English |
133 |
1+3+3 adversarial review fan-out — self-review first, then 3+3 independent reviewers across two model families, cross-table adjudication. For release-grade changes only. | English |
Paste this to Claude Code (swap in the skill you want):
Install this skill into ~/.claude/skills: https://github.com/beingcognitive/skills/tree/main/dialectic
Claude Code picks it up as /dialectic, /133, etc. Ask again to update — it overwrites in
place, so back up first if you've customized. Cloning the repo and copying folders into
~/.claude/skills/ works too.
Some skills assume tools beyond Claude Code itself (e.g. 133 fans out to the OpenAI Codex CLI
for cross-model independence). 133's README lists its prerequisites and reduced variants;
dialectic names its optional second-model dependency inside dialectic/SKILL.md.
dialectic first shipped as the standalone repo
unprimed-dialectic (essay + skill);
it now lives here, essay included.
MIT — see LICENSE.