You are a senior software architect. Review the coupling of this module, propose hard boundaries and a step-by-step migration plan...
The command-line
arsenal for AI engineers
Prompts that actually ship code. Configs that survive Mondays. Scripts that go in your dotfiles and stay there. No fluff, no hype — just output.
// prompts
The prompt matrix
Indexed by role × scenario, versioned, with example output.
Given this API signature, produce 5 equivalence-class tests + 3 boundary cases in pytest style with parametrize...
Analyze this query's plan on PostgreSQL 16, identify missing indexes and suggest a rewrite that preserves semantics...
Design 3 chaos engineering scenarios for this service. Include trigger commands and concrete rollback procedures...
Audit this component for WCAG AA. List every violation with the smallest patch that fixes it without regressing UX...
Clean, dedupe and aggregate this CSV, then emit a same-vs-anomaly report and a one-paragraph explanation...
// claude code
Configs that just work
Hand-picked .claude/settings.json, hooks, status-line snippets and slash commands — covering the workflows you actually do. Every config ships with a minimal runnable example.
- → Safe allow / deny matrix (never delete by accident)
- → Hook templates: lint-on-save + test-on-commit
- → Plan / 1M-context / Bedrock switching presets
{
"permissions": {
"allow": [
"Bash(pnpm:*)",
"Bash(git status:*)",
"Edit",
"Write"
],
"deny": ["Bash(rm -rf:*)"]
},
"model": "claude-opus-4-7",
"env": {
"CLAUDE_CODE_USE_BEDROCK": "0"
}
} // cli scripts
The CLI library
Organized by scenario. Each script is dogfooded on a real machine. POSIX shell, fish, and Python — grab what you need.
| Path | Script |
|---|---|
| git/ | auto-rebase.sh |
| docker/ | prune-safe.sh |
| k8s/ | ctx-switch.fish |
| ai/ | claude-pipe.sh |
| ai/ | prompt-bench.py |
| git/ | blame-blame.sh |