개발 스킬

Plan worthwhile local tasks by checking dependencies, effects, resources, and platform boundaries, then execute the exact approved plan. It can use safe parallel work or suitable Apple-silicon backends when the project exposes them.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Speed up independent local build or test steps
• Plan a task with several dependency-linked operations
• Choose an Apple-silicon backend for supported numerical or media work

Structure engineering work performed by AI agents: define completion criteria, split tasks into verifiable units, run evaluations before and after changes, and choose model tiers by complexity. Track effort and focus reviews on risks and behavior.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Breaking an implementation into verifiable tasks
• Setting evaluation and regression checks before coding
• Choosing a model tier for each engineering task

Assess a concrete long-running Python entrypoint to see where repeated work can safely run in parallel and whether the overhead is worthwhile. When source optimization is authorized, make a minimal refactor and require correctness and performance evidence before claiming an improvement.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Assessing a slow Python command for parallel work
• Checking whether loop iterations are independent
• Reviewing a proposed process or thread pool

Turn a change request into one actionable end-to-end slice with an observable result. Define an example that can be checked, identify the owners and callers in scope, specify what should fail now and pass after the change, and record unresolved choices or later work.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Scoping a change before implementation
• Turning a request into a checkable acceptance example
• Clarifying which callers and tests a change affects

Inspect Bethesda game plugins and handle record conflicts or compatibility patches through the xEdit automation daemon. It routes work through MCP tools, checks daemon capabilities, and emphasizes controlled changes with previews and readback verification.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Inspect records in a Bethesda game plugin
• Compare conflicting mod records
• Create a reversible compatibility patch

Review a code-debt registry and rank entries by churn and Fowler quadrant, a way to classify how a shortcut was made. Get a prioritized list first, then work through requested items one at a time.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Prioritizing code debt to address
• Reviewing specific debt entries
• Deferring stale or low-impact entries

Implement an accepted code or configuration change, or repair a known defect, with small scoped edits. Preserve existing test expectations, check live consumers, run relevant checks, and report exact commands and results, including checks that were not run.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Implementing an accepted code change
• Repairing a known defect without weakening tests
• Reporting checks and affected consumers

Rebuild an authorized website or page as a local implementation by inspecting its layout, assets, responsive states, and interactions. Record component specifications and screenshots, then validate the result visually in a browser with evidence artifacts.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Recreate an authorized website locally
• Migrate a page while preserving its appearance
• Study a page's layout and interactions

Clarify what domain terms mean in their specific contexts, using repository documentation, code, callers, types, and tests. Surface disagreements between intent and behavior, distinguish competing meanings with examples, and identify compatibility work needed for shared-name changes.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Resolving disagreement over what a domain term means
• Checking whether code behavior matches the intended vocabulary
• Assessing compatibility work for a proposed rename

Run a single prompt through Claude Code's headless print mode for scripts or automation. Set a time limit and allowed tools, then capture the process status and output so a caller can assess what happened.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Automating a single Claude Code prompt
• Running a read-only review with limited tools
• Capturing output from a bounded Claude run

Run a supplied task in headless AGY (Antigravity) when the caller explicitly requests it, then report the command, permission posture, model, deadline, outcome, and artifacts. It checks the live CLI options and never silently switches to another agent runtime.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Run a task in explicitly requested Antigravity
• Collect output and changed-file paths from an AGY run
• Check the current AGY CLI options and models

Handle explicitly routed backend work by inspecting a project's architecture and conventions, defining API or data contracts, implementing focused changes, and recording validation evidence. It covers areas such as persistence, migrations, jobs, security, and reliability.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Implement a backend API change with its contract and tests
• Plan a database migration and check rollout risks
• Review a background job or cache behavior

Use this skill to write or refresh the managed “Tech debt operations” section in AGENTS.md. It inserts cached quality commands and shared guidance for recording technical debt and drafting architecture decision records, while leaving the rest of the file untouched.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Adding shared debt management guidance to AGENTS.md
• Refreshing cached quality commands
• Recording architectural decisions alongside debt entries

Record deferred engineering decisions and temporary workarounds in a debt registry, with context such as effort, ongoing cost, affected area, and payoff trigger. Entries receive short labels so they can be reviewed or dropped later.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Logging a TODO or unfinished implementation
• Recording a temporary default or bypassed check
• Tracking a swallowed error or loosened type

Run a single prompt with headless Codex and capture its output and exit status as one traceable result. Useful for a one-off command or CI step when you need a bounded run and a clear record of what happened.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Running one Codex prompt in CI
• Capturing the output of a one-off Codex run
• Checking a result without automatic retries

Coordinate two agents as they independently research a design problem, discuss findings, and converge on a specification. The workflow includes drafting and review, then presents the plan for human approval before implementation begins.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Researching a feature design with a partner agent
• Comparing independent design findings
• Drafting and reviewing a shared specification

Guides installation of the Julia programming language runtime on macOS, Linux, or Windows. It covers the official installer and platform-specific archive fallback, plus adding Julia to PATH so it can be launched from a terminal.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Install Julia on a machine without the runtime
• Use a tarball when the installer is blocked
• Set PATH so Julia runs from a terminal

Create a Julia project environment for scripts that use external libraries, or set up a reusable package with its project metadata and module file. The skill covers the default in-place layout and an explicitly requested subdirectory layout.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Setting up dependencies for a Julia script
• Creating a reusable Julia package
• Choosing a package directory layout

Use this skill to inspect and manage Bitbucket repositories, pull requests, branches, issues, pipelines, permissions, and webhooks through bkt. It helps choose commands for Bitbucket Cloud or Data Center and return structured results for automation.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing pull requests and repository status
• Managing branches, commits, and Cloud issues
• Inspecting or running Cloud pipelines

Find the current stable Julia version from JuliaLang’s machine-readable release metadata, including the newest patch in a chosen minor series. The skill also shows fallback lookup methods and how to obtain platform download URLs and SHA-256 hashes for verification.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Updating a Julia version value in an install script
• Finding the newest patch in a specific Julia minor series
• Getting a platform-specific Julia download URL and SHA-256 hash

Sets up a Bethesda game modding workspace by detecting Mod Organizer 2, selecting game knowledge packs, and optionally preparing xEdit and the xtl translator CLI. It initializes project logs and checks that the configured tools can respond.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Preparing a new Bethesda modpack project
• Checking a first time MO2 setup
• Setting up xEdit for plugin editing

Compile Papyrus source files (`.psc`) into game bytecode (`.pex`), or decompile bytecode for inspection and patching. The skill also detects available Creation Kit tools and handles Starfield Guard syntax during decompilation workflows.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Compile a Papyrus source file for a supported game
• Decompile a PEX file to inspect its script
• Check whether the compiler and decompiler are installed

Within an active Codex Project Autopilot project, turn discovery or approval inputs into a scoped implementation plan. Compare minimum, balanced, and expanded options, recommend one in plain language, and record deliverables, acceptance criteria, risks, and needed manual steps.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Planning implementation after project discovery
• Comparing three scope and effort options
• Choosing a practical technology stack for an MVP

Build and query a local graph of code relationships to find connections, callers, architectural hubs, and likely change impact. It can reduce repeated repository navigation for large, tangled codebases; its benefit depends on project size and structure, and it does not replace reasoning about design or performance.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Mapping an unfamiliar codebase
• Finding what connects two code areas
• Checking callers and change impact