kpab

10 Agent Skills built specifically for Claude Fable 5 — intent, boundaries, and verification hooks instead of the step-by-step micromanagement older-model skills relied on. See skills/ for the individual skill files.

Choose and tune the effort setting for Claude Fable 5 workloads. Use whenever the user asks which effort level to use ("what effort should this batch job run at?", "is xhigh worth it here?"), complains that Fable 5 is slow or expensive, wants to cut latency or token cost, designs a pipeline/harness that mixes routine and hard tasks, or reviews API code that sets output_config.effort.

Stop Claude Fable 5 from over-planning, re-deriving settled facts, or surveying options it will never pursue. Use in any interactive or agentic session where responses feel slow, turns run long on simple asks, or the model keeps restating context before acting. Especially valuable at high effort settings and in ambiguous, multi-threaded requests.

Keep Claude Fable 5 running to completion in unattended pipelines — no stopping on a statement of intent, no mid-run permission questions, no self-imposed session splits over context worries. Use for overnight runs, scheduled jobs, CI agents, and any harness where nobody is watching to type "continue".

Make Claude Fable 5 progress reports verifiable against actual tool results during long autonomous runs. Use for any multi-hour or overnight agent session, scheduled pipelines, CI agents, or whenever the user has been burned by status updates claiming work that wasn't done. Apply before launch, not after a fabricated report appears.

A file-based lesson memory that Claude Fable 5 reads and maintains across sessions. Use for any recurring agent (daily jobs, long projects, team assistants) where the same corrections keep being re-made, and when bootstrapping a new agent from past session history. Fable 5 benefits from recorded lessons noticeably more than prior models did.

Keep Claude Fable 5 changes minimal — no unrequested refactors, abstractions, defensive code, or feature creep. Use in code review, bug fixes, and any coding session where diffs come back bigger than the ask, with extra helpers, error handling, or "while I was here" cleanup. Particularly important at high and xhigh effort.

Make Claude Fable 5's final report readable to someone who saw none of the work. Use for long agentic sessions, overnight runs, and multi-tool tasks — whenever summaries come back as arrow-chain shorthand, invented abbreviations, or references to reasoning the reader never saw. Pair with grounded-progress for status content; this skill governs the prose.

Provides cryptography attack techniques for CTF challenges. Use when attacking encryption, hashing, signatures, ZKP, PRNG, or mathematical crypto problems involving RSA, AES, ECC, lattices, LWE, CVP, number theory, Coppersmith, Pollard, Wiener, padding oracle, GCM, key derivation, or stream/block cipher weaknesses.

Provides binary exploitation (pwn) techniques for CTF challenges. Use when exploiting buffer overflows, format strings, heap vulnerabilities (House of Orange, Spirit, Lore, Apple 2, Einherjar, tcache stashing unlink), race conditions, kernel bugs, ROP chains, ret2libc, ret2dlresolve, shellcode, GOT overwrite, use-after-free, seccomp bypass, FSOP, stack pivot, sandbox escape, Windows SEH overwrite, VirtualAlloc ROP, SeDebugPrivilege escalation, or Linux kernel exploitation (modprobe_path, tty_struct, userfaultfd, KASLR bypass, SLUB heap spray).

Provides web exploitation techniques for CTF challenges. Use when solving web security challenges involving XSS, SQLi, SSTI, SSRF, CSRF, XXE, file upload bypasses, JWT attacks, prototype pollution, path traversal, command injection, LaTeX injection, request smuggling, DOM clobbering, Web3/blockchain, authentication bypass, SAML exploitation, OAuth/OIDC, open redirect chains, subdomain takeover, or CI/CD credential theft.

包括的なSEO最適化スキル（日本語版）。Googleの公式ガイドラインに基づく技術SEO、コンテンツSEO、構造化データ（JSON-LD）、Core Web Vitals、E-E-A-T対策を網羅し、実践的なコード生成とサイト監査ワークフローを提供。SEO、検索順位、メタタグ、robots.txt、サイトマップ、canonical・hreflang、schema.org構造化データ、リッチリザルト、LCP/INP/CLS、Lighthouse・PageSpeedスコアに関する相談や、サイト監査の依頼があったときに使用する。

Comprehensive SEO optimization skill based on Google's official guidelines. Covers technical SEO, content SEO, structured data (JSON-LD), Core Web Vitals, E-E-A-T strategies, practical code generation, and site audit workflows. Use when the user asks about SEO, search rankings, meta tags, robots.txt, sitemaps, canonical or hreflang tags, schema.org structured data, rich results, LCP/INP/CLS, Lighthouse or PageSpeed scores, or requests a site audit.

デモサイトを Cloudflare Pages にデプロイする。「デプロイして」「デモを公開して」と言われたら使う。

実装をデザインハンドオフ（docs/design/）と照合し、色・タイポ・余白・シャドウ等の乖離を洗い出す。「デザイン確認」「デザインと合ってる？」「design check」と言われたら使う。

ネタ系スキルをリストに1件追加する。「このスキル追加して」「これ載せて」とURLを渡されたら使う。skills.yaml への追記と README 再生成までを一貫して行う。掲載対象は GitHub リポジトリのみ。

ネタ系スキルの新規候補を発掘する。「新しいネタスキル探して」「候補集めて」「巡回して」と言われたら使う。GitHub を定型クエリで巡回し、未掲載のリポジトリ候補をリストアップする。

Brutally roast the user's uncommitted changes — every insult maps to a severity-ranked, concrete fix. Use when the user says /roast, "roast my code", "roast this diff", or asks for a merciless / brutal code review.

Use this skill when someone wants a first-impression, outsider's audit of a GitHub repo to figure out why it isn't earning stars or attention, or wants a prioritized plan before a launch. Trigger on framings like "review my repo like a stranger," "why isn't this getting stars," "what's wrong with our repo / why does it look unprofessional," "we're about to launch/release, what should we fix first," or "nobody's starring this even though it's useful" — even without the words "star" or "growth." Also trigger on explicit growth language: more stars, スター, 伸ばす, promote, discoverability, popularity, visibility. Accepts a GitHub URL or a local repo path, including a bare "how can I improve this?" with a link. Produces a full prioritized audit report (Quick Wins / High-Impact / Later), not a single fix. Do NOT use for narrow single-task asks: fixing README typos, writing a CONTRIBUTING.md, picking a license, reviewing a specific PR, refactoring folder structure, or fetching star counts via the API.

Frostpane のテーマを再生成して見た目を確認するときに使う。「テーマをプレビュー」「色を変えたので確認したい」「再生成して」などと言われたら実行する。

A Socratic sparring partner that helps the user understand a difficult text, paper, or concept well enough to explain it in their own words. It does not summarize or ghost-write; it makes the user reconstruct the logic and pushes back on their wording. Use it whenever the user wants to understand, grok, or internalize something for themselves — e.g. "what does this mean", "I want to understand this myself", "help me make this click", "I want to organize this for my own use", "break this down for me", "I want to check that I really got it", or in Japanese 「これどういうこと」「自分のために理解したい」「腹落ちさせたい」「自分用にまとめたい」「噛み砕きたい」「ちゃんと分かったか確かめたい」. Also applies when working through papers, articles, hard concepts, or jargon. Do NOT use it when the user clearly wants a deliverable written for them ("summarize this", "write an article", "draft an explanation" / 「要約して」「記事にして」「説明文を書いて」) — in that case skip the method and confirm the goal (understanding vs. a deliverable) exactly once.