Personal Claude Code skills and reference material.
| Skill | Path | Description |
|---|---|---|
fix-simple-issues-implementer |
agent-implementer/ |
Implement a single labeled GitHub issue in an isolated worktree, review-ready |
agent-diff-reviewer |
agent-diff-reviewer/ |
Haiku sub-agent that reviews an implementer agent's diff before a human sees it |
codebase-analysis |
code-analysis/prod/ |
Production-grade review for Bun/React/TanStack/Zustand/Drizzle/PostgreSQL stack |
custom-analysis |
code-analysis/custom/ |
User-defined codebase analysis against any criteria or lens |
grilling-designs |
grill/ |
Stress-test a plan or design by relentless questioning until shared understanding |
laravel-best-practices |
laravel/best-practices/ |
Laravel 12 — use what the framework already provides |
project-ideas |
project-ideas/ |
Generate original project ideas tailored to Thomas's stack |
prompt-improver |
prompt-improver/ |
Rewrite a draft AI prompt using prompt-engineering best practices |
schematic-writer |
schematics/ |
Architecture schematics, call graphs, and data flow docs |
skill-improver |
skill-improver/ |
Audit and rewrite this repo's own skills against agentskills.io best practices |
think-twice |
think-twice/ |
Pause before expensive work to find the cheaper path |
thomas-learning |
thomas-learning/ |
Concept explanations pitched at intermediate-to-senior level |
Each skill has a concise SKILL.md core (under 100 lines) with extended detail in references/ files loaded on demand.
agent-implementer/
SKILL.md # fix-simple-issues-implementer: worktree → implement → validate → log → review
agent-diff-reviewer/
SKILL.md # Read-only diff review, spawned by agent-implementer step 5
code-analysis/
custom/
SKILL.md # Custom criteria analysis (any stack)
references/playbooks.md # Grep commands per audit type
prod/
SKILL.md # Production analysis (Bun/React/Drizzle/Zustand)
references/anti-patterns.md
references/solid.md
references/structure.md
grill/
SKILL.md # Design stress-tester
laravel/
best-practices/
SKILL.md # Laravel 12 built-in features reference
references/ # Eloquent, validation, security, queues, mail, notifications
project-ideas/
SKILL.md # Project idea generator
prompt-improver/
SKILL.md # AI prompt rewriter
references/principles.md # Full prompting-principle catalog with before/after pairs
schematics/
SKILL.md # Architecture schematic writer
skill-improver/
SKILL.md # Audits/rewrites skills in this repo for efficiency and spec compliance
think-twice/
SKILL.md # Cheaper-path check before heavy/repetitive work
thomas-learning/
SKILL.md # Personal learning skill
Run the scaffold script — it handles folder creation, frontmatter, validation, and listing in one step:
./scripts/new-skill.sh <domain> <folder> <name> <description>Pass . as <domain> for a flat, top-level skill (no subdomain), e.g. ./scripts/new-skill.sh . my-skill my-skill "...".
Or run it with no arguments for interactive prompts. After it runs, open the generated SKILL.md and fill in the skill content.
| Script | Purpose |
|---|---|
./scripts/new-skill.sh <domain> <folder> <name> <desc> |
Scaffold a new skill — validates name format, creates SKILL.md, optionally creates references/ |
./scripts/validate-skills.sh |
Spec compliance check — name format, description length, line count, placeholders. Runs on every commit. |
./scripts/test-skill.sh <skill-dir> |
Quality check — description trigger keywords, referenced files exist, progressive disclosure, body content |
./scripts/list-skills.sh |
Print all registered skills with name and path |
# 1. Scaffold
./scripts/new-skill.sh laravel hooks laravel-hooks "Trigger on Laravel lifecycle hook questions"
# 2. Fill in the SKILL.md body
# 3. Quality check
./scripts/test-skill.sh laravel/hooks
# 4. Spec compliance (also runs automatically on commit)
./scripts/validate-skills.shnew-skill.sh accepts all four arguments positionally, or runs interactively with no arguments.
Use the skill-improver skill directly in conversation ("improve the X skill", "audit my skills") — it runs
validate-skills.sh and test-skill.sh against the target, checks it against the agentskills.io
checklist, and applies fixes.