개발 스킬

Prepare updates to a local code index, a map used to navigate a repository. Summarize structure and task-relevant context for handoff, while distinguishing planned commands from verified outputs and generated files.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Refreshing an outdated code index
• Summarizing repository structure before planning
• Preparing task-specific files and entry points for handoff

Plan a behavior-preserving refactor of oversized UI components, addressing state management before hooks, component extraction, and naming. Preview changes with safety reasons, then hand off approved implementation with tests that check behavior before and after.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Planning how to split an oversized component
• Untangling state managed by many boolean flags
• Reviewing excessive useEffect chains

Examine a technical proposal through independent perspectives, challenge their findings, and resolve objections. The result is a structured set of constraints, decisions, risks, and open questions for a separate planning workflow.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing a technical proposal before planning
• Challenging assumptions behind a migration
• Organizing risks and unresolved questions

Analyze failed builds and CI checks, including automated tests, using fresh logs. Group failures by likely cause, distinguish newly introduced issues from existing ones, and prepare a minimal repair handoff with ordered verification steps.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Investigating a failed build or compile error
• Sorting type check, lint, and test failures
• Distinguishing new failures from existing problems

Compare Codex, Claude Code, Hermes coding, and oh-my runtimes before assigning development work. Identify missing tools or credentials, clarify how work will be handed over, and distinguish preparation from confirmed execution.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Choosing between Codex and Claude Code
• Checking tools and credentials before handing over work
• Reviewing readiness after changing execution environments

Investigate incorrect application behavior by recording a reproduction, comparing competing explanations, and choosing low-cost observations that distinguish them. Prepare a fix only after evidence identifies the cause, carrying the reproduction forward as a regression test.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Investigating tests that fail intermittently
• Finding the cause of an incorrect return value
• Diagnosing lost updates during concurrent processing

Turn a repository's actual code structure into one readable PlantUML diagram and rendered PNG. Show packages or modules, public elements imported elsewhere, and dependency links, with a visible legend explaining which details were omitted to keep the picture readable.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Introducing repository structure to a new teammate
• Exploring dependencies around a specific package
• Adding a code structure diagram to a PR or document

Draft commit messages and pull request descriptions using the repository’s existing conventions. Explain why the change matters and separate commands with observed execution records from planned or skipped checks, so validation claims match the available evidence.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Drafting a commit message for a code change
• Writing a PR description using the repository template
• Separating executed tests from skipped checks

Plan server, API, and data changes before implementation. Define who can access each operation, consistent response formats, error handling, and database migration steps, with compatibility plans for existing consumers and rollback points for storage changes.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Design an API and its database schema before coding
• Clarify authentication and authorization checks
• Standardize success and error response formats

Review pull requests, code changes, or documented claims for bugs using concrete evidence. Receive prioritized findings, a separate assessment of whether requirements are met, and clear lists of checked areas, unanswered questions, and testing gaps.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Find bugs in a pull request before release
• Check changes against stated requirements
• Compare README claims with supporting evidence

Explore an unfamiliar code repository and build a practical guide to its structure, reading order, terminology, risks, and unknowns. Use observed files to suggest low-risk starter tasks and the commands needed to verify them.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Getting oriented in an unfamiliar repository
• Choosing which modules and tests to read first
• Preparing repository guidance for a new contributor

Capture reusable lessons from agent work, including user corrections, recurring problems, and useful methods. Keep observations for weekly reviews that inform skill updates, and identify repeated workflows that could become new skills.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Preserve reusable lessons from user corrections
• Review accumulated skill improvements weekly
• Identify repeated workflows as new skill candidates

Reviews generated skills against the Ars Contexta knowledge base and proposes improvements with research support. Checks existing customizations, explains risks and workflow changes, and applies only the updates the user approves.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing generated skills for methodology improvements
• Comparing customizations with recommended changes
• Choosing updates based on research support and risk

Review openai-go pull requests, branches, or commit ranges for Go behavior, SDK compatibility, generated-code ownership, and security. Get prioritized, actionable findings with evidence, suggested fixes, completed checks, and clearly stated review gaps.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing openai-go pull requests
• Checking SDK compatibility across commits
• Reviewing credential and provider changes

Audit openai-go during recurring maintenance to identify correctness, security, performance, and design issues. Receive at most one evidence-backed recommendation, with compatibility considerations, generated-code ownership, validation requirements, and the next area to review.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Recurring openai-go maintenance audits
• Rotating security and architecture reviews
• Selecting one evidence-backed improvement

Explain a team's shared Claude Code setup to new teammates. Summarize what each configuration file controls, distinguish shared and personal settings, and identify what the team-kit plugin adds without changing files.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Onboarding a teammate after cloning the repository
• Understanding shared versus personal Claude Code settings
• Checking what the team-kit plugin adds

Turn a repeated workflow into an agent skill with SKILL.md and supporting reference files. Plan its structure, improve existing instructions, and assess quality using a ten-part rubric with scores and prioritized recommendations.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Turning a repeated manual workflow into a skill
• Planning and writing a new agent skill
• Fixing instructions in an existing skill

Try a simple skill invocation with `/example`, which returns a friendly greeting and explains that the demo was loaded from the workspace skills directory. The sample also shows the basic file structure for creating your own skills.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Trying a skill through the /example command
• Checking the demo’s greeting and explanation
• Learning the basic structure of a custom skill

Build interface prototypes and changes using an existing app’s structure, components, and design rules. Check relevant screen states and interactions, then prepare current screenshots, test results, and outstanding checks for review.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Prototype screens using existing app components
• Check light, dark, loading, and error states
• Prepare screenshots and test results for review

Start an engineering handoff by clarifying the project, repository rules, and allowed actions. Summarize a selected branch or PR/MR, a proposed code change, including its checks, reviews, delivery status, and most important next step.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Follow up on an ongoing PR or MR
• Break a requirement into reviewable changes
• Check whether a branch is ready for delivery

Carry engineering tasks from scope definition through implementation and validation. Preserve existing work and prepare a reviewable delivery with code changes, current check results, unresolved questions, and reproduction materials.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Defining deliverables and acceptance checks for a development task
• Implementing changes while preserving existing work
• Preparing code changes and validation evidence for review

Create, update, debug, and package Cindy plugins using the host’s authoring references. Define behavior, interfaces, and permissions, validate the implementation, and produce a .cindy package with Forge tools, with installation or publication handled when requested.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Build a new Cindy plugin from a feature request
• Update or debug an existing Cindy plugin
• Check host API usage and permission declarations

Create and edit reusable Cindy Skills with focused instructions and supporting files. It also covers searching public or organizational Skills on SkillHub, uploading local Skills, and publishing new versions of Skills you authored.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Create a reusable Skill for recurring work
• Simplify instructions in an existing Skill
• Find public or organizational Skills on SkillHub

Run coding agents from scripts, continue saved conversations, submit queued tasks, and compare responses across agents. Use structured JSON output to collect results and build workflows that coordinate multiple agents.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Continue an agent conversation across commands
• Compare responses from multiple coding agents
• Collect agent results as JSON in scripts