gapp
claudeApp-building harness — idea intake to App Store submission, stage by stage
12 skills
Core productivity tools — base plugins
Plugin gbase from marketplace godsenal in Godsenal/godsenal-cc. Commands read from the repo's manifest — yesterday.
This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.
blast-radiusFinds what a change could break somewhere else before it ships — beyond the diff, where grep stops — and proves the one fact it's safe because of by running real code instead of writing it up. Use for 'blast radius of X', 'what could this break', '이거 뭐 깨질 수 있어', or a small diff you don't trust yet.
bug-fixFixes a bug at its root cause, not where it surfaced — reproduces it first (via gbase:reproduce), narrows the cause by ruling out hypotheses with runtime evidence, asks whether the triggering state is legitimate before adding any guard, fixes every instance of the pattern, and verifies on the same s
compat-checkScans a code change for backward-compatibility hazards and gives the safe deploy/script order. Reads a diff (uncommitted, a commit range, or a PR) and flags migrations, API/contract changes, queue/event schema changes, new required env/secrets, feature flags, and backfill scripts — then emits an ord
correctFinds the mistakes agents keep repeating in this repo and makes each one impossible: architecture first, then types, then a lint/CI check whose error names the fix, then a test, with REVIEW.md text last. Proves each check fails on a real past mistake. With --session it runs as a retrospective on one
create-verifyGenerates a project-local verification skill (.claude/skills/verify-<app>/) that drives the real app the way a user does — web, mobile simulator, CLI, or API — and captures evidence, plus a per-feature map; proves the generated skill by running it once before handing it over. With --maintain, audits
design-tasteHolds the user's own UI/UX taste, collected from their design feedback across apps, as rules to apply while working: let layout, defaults, and affordances lead the action instead of a caption that explains the UI (\"필터는 아래에서 고를 수 있어요\"), interactive onboarding instead of instruction screens, no feat
ego-profileego-browser에서 **다른 계정/프로필로** 작업해야 할 때 쓴다. task space는 기본 프로필로 열리므로 회사↔개인 계정이 갈린 서비스(PostHog, GitHub, Google, AWS 등)에서 "권한이 없다 / 조직이 안 보인다 / 프로젝트가 목록에 없다"가 나온다. 이때 로그아웃·재로그인으로 헤매지 말고 여기 적힌 `listProfiles()` + `newTaskSpace(name, profileId)` 를 쓴다. "다른 계정으로", "회사 계정 말고", "프로필 바꿔서", "계정 전환" 같은 말이 나오거나, e
goShips a finished change end to end — reviews and polishes it in parallel, applies verified review fixes, runs the project's fast checks, opens the PR via gbase:branch-pr, and watches it to merge via gbase:monitor. Review happens before the PR so fixes land in the initial commits. Use when the user s
howExplains how a part of the codebase works — subsystem architecture, runtime flow, where things live — at the level a senior engineer needs to start working in it, citing real files and functions. Also answers placement and ownership questions ('where should this live', 'which module owns this', 'is
jevPuts TypeSafe's Jev (a System One decision model: Choice, Score, Noul, Box answers with calibrated probabilities, about $0.04 per million input tokens, no text generation) into the user's bots and apps where a cheap, typed judgment beats a regex, a keyword list, or an LLM prompt that returns JSON —
karpathyMerges Andrej Karpathy's behavioral coding-agent guidelines (Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution) into this project's CLAUDE.md — additively and idempotently, never clobbering project-specific rules. Use when the user asks to add the Karpathy rules/skills,
mockupTurns a design request into a quick, clickable HTML mockup instead of a description — one self-contained HTML file at real device size, built from the app's actual screens, colors, fonts, and copy, with several options compared on one page and PNG screenshots of each step, delivered where the user c
monitorWatches a PR until it merges or closes — auto-fixes clear CI failures, applies clearly-requested review changes, resolves safe merge conflicts, replies on every review comment it handles, surfaces testable app/preview links, and asks about anything ambiguous. On a PR that wasn't already reviewed it
pauseStops in-flight work cleanly so a cold-start agent can resume it — finishes or backs out of the current step, commits the tree as one wip commit whose body is the resume note (intent, progress, what's verified, next step, gotchas), and pushes nothing new. Use only when explicitly asked to pause, sto
pickupResumes or takes over in-flight work — from a gbase:pause wip commit, a prior Claude Code session's transcript, another agent's pushed branch, or an open PR — by reading the trail as authoritative instead of redoing it, naming the resume point, and verifying inherited claims on the real artifact bef
polishPolishes the current diff without changing behavior, in two passes — a three-lens cleanup (AI cruft & dead code, reuse & duplication, clarity & efficiency) run by parallel sub-agents, then a structural pass that collapses needless helpers and abstractions. Use after writing a non-trivial change, whe
reproduceTurns a bug report or an improvement request into an executable check that fails right now — a test, script, HTTP request, CLI run, or verify-skill journey that shows the bug, or a measurement of the requested criterion at its current value — and returns the command, its failing output quoted, and t
reviewAdversarially reviews the current change for defects — bugs, regressions, security holes, missing handling. Delegates finding to the built-in code-review, has independent skeptic sub-agents try to refute each finding, auto-fixes the clear and local survivors, and reports the rest. Local-only; never
setupPrepares a repo for the gbase ship loop — records the project's own verify commands in CLAUDE.md or AGENTS.md (after running each once on current code), seeds a short REVIEW.md that gbase:review follows, adds a formatter hook, and offers a project-local verification skill. Idempotent; writes only af
unslopRewrites prose to remove AI writing tells — filler, hedging, AI vocabulary, mannered metaphors, em dashes, inline-header lists, chatbot phrases, and Korean 번역투 — while keeping the meaning and tone. Use on PR titles and bodies, commit messages, docs, READMEs, release notes, Slack drafts, or a reply b
whyInvestigates why code is the way it is — design rationale, a tradeoff, a regression's origin, a magic number, defensive code — by anchoring on git blame and PR history, then querying every connected evidence source (issue tracker, docs, chat, error tracking, observability, analytics MCPs) in paralle
workHow to carry out a non-trivial task — classifies the request (investigation, bug, feature, refactor, perf, prototype, or run-until-done), copies that route's steps into the todo list, and works them under a few core principles with gbase's scout/worker subagents. Use before starting a task that touc
/gbase:branch-prAutomate branch creation, commit grouping, and PR generation from current changes
/gbase:find-skillsFind skills that match your project's tech stack
scoutRead-only gbase scout on Sonnet. Use PROACTIVELY when answering means reading more than a few files, or a log, CI output, diff, or transcript would flood the main thread — it sweeps them and returns the conclusion with path:line evidence, not the dump. Has no edit tools and is told to change nothing
workergbase code-change delegate on the session's model. Use for one self-contained unit of implementation — a feature slice, a refactor step, a fix whose cause is already known, a mechanical sweep — briefed with the unit, its check, and file pointers. Proves the change on the real artifact and reports a
sh "${CLAUDE_PLUGIN_ROOT}/hooks/session-start.sh"
App-building harness — idea intake to App Store submission, stage by stage
12 skills
Korean content harness — faceless short-form videos, Instagram card news, Naver blog strategy, Threads/X posts, and comments: common rules, feedback intake, and one skill per format
10 skills
Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).
39 skills
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
15 skills · hooks
Official Next.js skills: adopt and optimize Cache Components, adopt Partial Prefetching, and verify runtime behavior against a running dev server.
6 skills
Talk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
Persistent memory for Claude Code: the next session starts where the last one left off.
29 skills · 1 MCP server · hooks
tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.
1 skill
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.