tomwangowa

tomwangowa/agent-skills

Hook Claude Code

Runs before the agent uses a tool for Bash tool calls, running an inline shell check. From tomwangowa/agent-skills.

tomwangowa/agent-skills

Settings file Claude Code

Agent settings declaring 1 hook event (PreToolUse).

tomwangowa/agent-skills

Instructions file

Claude Code instructions for tomwangowa/agent-skills, covering claude.md, repository overview, skill structure, skill.md format and available skills.

tomwangowa/agent-skills

Skill Claude CodeCodex

Produce weekly, daily, or monthly sharing-prep AI news deep-analysis reports for TrendLife AI Taskforce. Use when asked for "AI 週報", "AI 日報", "weekly insight", "daily insight", "AI sharing prep", "sharing session", or AI news analysis.

tomwangowa/agent-skills

Skill Claude CodeCodex

Find, deeply understand, and produce engineer-friendly digests of arXiv AI papers for TrendLife AI Taskforce meeting sharing. Use when asked for "arXiv 導讀", "論文導讀", "paper digest", "arxiv digest", or given an arXiv URL to analyze. Supports single paper deep-dive, topic search with candidate selection, and multi-paper…

tomwangowa/agent-skills

Skill Claude CodeCodex

Systematically extract explicit and implicit assumptions from technical documents, design plans, or feasibility assessments. Classifies each assumption by visibility (explicit/implicit), verification status (verified/unverified/falsified), recommended verification method, and risk impact. Use when reviewing any…

tomwangowa/agent-skills

Skill Claude CodeCodex

Use when Codex needs to turn a vague idea into an approved design before implementation, or when a user wants to grill / stress-test a plan or idea (triggers: grill, 烤, stress-test). Apply to new features, architecture changes, refactoring, and workflow design. Interview one decision at a time with a recommendation…

tomwangowa/agent-skills

Skill Claude CodeCodex

Use this skill when the user is designing a Claude Code workflow, choosing between mechanisms (hook/skill/subagent/MCP/CLAUDE.md), or packaging automation for team distribution. Also triggered by the /sos slash command.

tomwangowa/agent-skills

Skill Claude CodeCodex

Use when Claude Code receives a code-review request, 'review', 'code review', 'quick review', 'native review', or a fast code check. Runs native (< 30 s) line-by-line analysis with adversarial pass, assumptions list, syntax-checker-verified syntax findings, and an optional ready-to-apply Refactored Patch. In Codex…

tomwangowa/agent-skills

Skill Claude CodeCodex

Use when the active agent is Codex and the user asks for code review, "review", "quick review", "check my changes", "看一下 code", "掃一眼", or when a Codex session reaches pre-commit auto-review. This Codex-native counterpart to code-review-claude performs structured local review with findings first, adversarial checks…

tomwangowa/agent-skills

Skill Claude CodeCodex

Use when a legacy Gemini review invocation needs a migration response. Explain the retirement; do not invoke Gemini or send code externally. Route generic review to code-review-claude in Claude Code or code-review-codex in Codex.

tomwangowa/agent-skills

Skill Claude CodeCodex

Analyze git history to tell the story of how code evolved. Use this Skill when the user asks to understand code history, explain evolution, show file timeline, or understand design decisions behind code.

tomwangowa/agent-skills

Skill Claude CodeCodex

Claims-first codebase audit that extracts documentation claims and verifies them against code. Use when asked to "audit", "verify docs match code", "check if README claims are true", or "validate documentation accuracy". Falsification-first approach.

tomwangowa/agent-skills

Command

Stuck On Setup — guide me through designing a Claude Code workflow.

tomwangowa/agent-skills

Skill Claude CodeCodex

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always.

tomwangowa/agent-skills

Skill Claude CodeCodex

Falsification-first research skill that actively seeks counter-evidence before supporting evidence to eliminate cognitive biases and ensure rigorous, objective conclusions.

tomwangowa/agent-skills

Skill Claude CodeCodex

Use when Tom wants to rewrite HIS OWN outward-facing content (簡報稿 / blog / newsletter / 貼文 / sharing slides) to strip AI-flavored writing, make it plainer (白話), and match his personal voice. Triggers — "去 AI 味", "去AI味", "幫我去 AI 味", "白話化", "改成我的口吻", "潤成我的語氣", "這篇 AI 味很重", "de-AI this", "rewrite in my voice", "make this…

tomwangowa/agent-skills

Skill Claude CodeCodex

Systematic knowledge extraction from document sets. Identifies core mental models, expert disagreements, knowledge gaps, and teachable frameworks. Use when given documents/URLs/notes to deeply understand, not just summarize. Triggered by "deep read", "help me understand this", "extract insights", "what are the key…

tomwangowa/agent-skills

Skill Claude CodeCodex

Use when you want a clear, at-a-glance view of git state in the Agent CLI — "git status" (enhanced), "git 狀態", "gst", "掃一下 git", "repo 狀態", "哪個 repo 還沒乾淨". Renders a box-drawing panel of the current repo (branch, ahead/behind, staged/modified/untracked, stash, in-progress ops) or, from a parent directory, a…

tomwangowa/agent-skills

Skill Claude CodeCodex

Use when ending a work session in any git repo, or when the user says "/handoff", "收工", "交接", "留交接", "hand off", "下班", or asks to produce a handoff for the next session. Auto-drafts a handoff (task, state, premises to re-verify, non-negotiable rules, top risk, next step, verified vs not-verified), shows it for review…

tomwangowa/agent-skills

Skill Claude CodeCodex

Generate interactive presentation from an outline or brief. Creates markdown-based presentations (Marp/Slidev/reveal.js) or standalone HTML presentations. Use when user asks to create slides, make a presentation, or generate a deck from an outline.

tomwangowa/agent-skills

Skill Claude CodeCodex

Translate markdown files to English and Traditional Chinese. Automatically detects source language and generates both -en.md and -zh.md versions. Preserves markdown formatting, code blocks, frontmatter, and links.

tomwangowa/agent-skills

Skill Claude CodeCodex

Empirically validate technical assumptions through minimal code experiments (5-30 min spikes). Use when a tech-feasibility assessment or design document contains claims that can be proven or disproven by running actual code. Triggered by "validate this assumption", "can X actually do Y?", "prove it with code", or…

tomwangowa/agent-skills

Skill Claude CodeCodex

Audit external narratives (articles, posts, marketing copy) against primary sources, or speak on behalf of the user as their AI proxy. Use when asked to "fact-check", "verify this article", "audit this claim", "help me respond", "speak for me", or "supplement context".