ThomasLucking

ThomasLucking/agents

Instructions file

Claude Code instructions for ThomasLucking/agents, covering agents — claude code skills repo, structure, adding a new skill, scripts and skill creation workflow.

ThomasLucking/agents

Skill Claude CodeCodex

Reviews code an agent just wrote in an issue worktree and reports findings back to the caller. Runs as a spawned Haiku subagent; used by fix-simple-issues-implementer after step 4, or on request to "review what you just changed".

ThomasLucking/agents

Skill Claude CodeCodex

Implements a single, well-scoped GitHub issue in an isolated worktree, validates it, logs it for human review, and has a Haiku agent review the diff. Trigger on "fix issue #42", "implement issue 17", or an issue labeled "Agent". Never commit or open a PR.

ThomasLucking/agents

Skill Claude CodeCodex

Senior code reviewer that evaluates changes across five dimensions — correctness, readability, architecture, security, and performance. Use for thorough code review before merge.

ThomasLucking/agents

Skill Claude CodeCodex

Interviews the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

ThomasLucking/agents

Skill Claude CodeCodex

Trigger when writing, reviewing, or debugging Laravel code. Enforces framework conventions: thin controllers, Form Requests for validation, Eloquent over raw queries, and proper use of built-in features. Also trigger on questions about Laravel architecture or "how do I do X in Laravel".

ThomasLucking/agents

Skill Claude CodeCodex

Generates original project ideas tailored to Thomas's stack. Trigger when Thomas asks for project ideas, wants something to build, says "I don't know what to build", "suggest a project", "what should I build next", or wants to practice a specific technology.

ThomasLucking/agents

Skill Claude CodeCodex

Rewrite a user's draft prompt into a stronger, clearer prompt for an AI model using prompt-engineering best practices (natural phrasing, explicit context, response directives, output format stated last, interrogative framing, few-shot examples, chain-of-thought, negative prompting). Runs a short interview to pull out…

ThomasLucking/agents

Skill Claude CodeCodex

Audits and rewrites one or more existing skills in this repo to maximize efficiency: trims bloat, tightens the description for reliable triggering, adds missing gotchas, and restructures oversized SKILL.md files into references/ for proper progressive disclosure. Trigger on "improve this skill", "clean up my skills"…

ThomasLucking/agents

Skill Claude CodeCodex

Pauses before expensive work to find the cheaper path. Use this skill before heavy computation, large code generation, repetitive data writing, or building more than what's needed — even if efficiency isn't mentioned explicitly. The goal is productive laziness: do less if less is enough.

ThomasLucking/agents

Skill Claude CodeCodex

Personal learning skill for Thomas. Activate when Thomas explicitly asks to learn something or wants an explanation: "explain X to me", "how does X work", "what is X", "I don't understand X", "teach me X". DO NOT activate for code tasks, debugging, or building things — only for learning and understanding.