개발 스킬

Review an AGENTS.md line by line to identify lasting principles, outdated mechanics, misplaced details, and repetition. When explicitly asked to rewrite it, produce a cleaner file and reread it from a newcomer’s perspective.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing an AGENTS.md file
• Finding project details that belong elsewhere
• Removing repeated or outdated guidance

Use Claude Code as an independent subagent when asked for Claude’s perspective or to delegate a clearly scoped task. Set the model and effort level, require evidence for findings, and verify its output before using it.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Requesting Claude’s second opinion on code
• Delegating a clearly scoped task to Claude
• Choosing Claude’s model and effort level

Turn a large or risky feature into small, reviewable implementation slices. The resulting spec defines boundaries, checks, and visible checkpoints so a team can verify progress before the whole feature is complete.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Breaking a risky feature into verifiable steps
• Planning browser or visual checkpoints
• Researching unfamiliar implementation approaches

Run an ordered closeout review for a substantial code change: clean up its structure, inspect the settled diff, then update and check related documentation links. Report findings and resolved or remaining issues across all three passes.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing a feature before calling it done
• Preparing a change before committing
• Cleaning up and documenting a code change

Reshape code when names become misleading, responsibilities overlap, or modules grow too broad. Find each concept’s real owner, remove redundant inputs and wrappers, and keep only compatibility paths tied to a concrete consumer and behavior.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Cleaning up stale names after a scope change
• Consolidating duplicated concepts and adapters
• Splitting an overburdened module by responsibility

Implement an existing feature spec through focused, reviewable passes. The skill tracks progress, verifies each contract, maintains the handoff, and periodically cleans up the plan and code so long-running work stays aligned through completion.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Implementing a multi-slice feature spec
• Keeping a long implementation plan current
• Verifying and reviewing each committed pass

Coordinate implementation of a specification with Codex writing the code. The skill keeps planning, integration, verification, review, and commit decisions with the orchestrator, while assigning implementation passes to Codex and checking the merged result.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Running implement-spec with Codex as implementer
• Delegating separate implementation slices to Codex
• Integrating and reviewing Codex implementation results

Update a mid-implementation spec with progress, decisions, preferences, unresolved issues, and machine-specific work, so a new agent or machine can continue from the repository without relying on the previous conversation. Then push the handoff references to the remote.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Handing an in-progress spec to another agent
• Moving implementation work to another machine
• Recording decisions and unresolved issues

Guide a step-by-step conversation to map what is known and unknown about a task across four quadrants. The completed map records decisions and open questions, helping clarify an ambiguous request or inform a later specification.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Clarifying an underspecified request
• Mapping unknowns before writing a spec
• Understanding a reference implementation before porting

Investigate performance problems by tracing production paths, estimating how work grows, and checking whether retries make progress. Rank findings by user impact and evidence, then guide small fixes that bound resource use while preserving recovery.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Investigating freezes, server errors, or memory exhaustion
• Finding retries or polling loops that make no progress
• Ranking a performance backlog with evidence

Review the design and implementation decisions an AI agent made beyond the original request. The skill traces the work, records each choice with its rationale, future impact, verdict, and confidence, and leaves the code unchanged.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Review AI-made decisions before merging or committing
• Audit a delegated agent’s work
• Check whether a working fix addresses the general cause

Analyze requests to Web APIs you own or are authorized to use, then build and verify data collection programs in Node.js or Python. The workflow can examine browser traffic and signatures, and supports pagination, resuming interrupted runs, and regression checks after interface changes.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Collecting data from an authorized Web API
• Investigating request signatures or obfuscated JavaScript
• Resuming a paginated collection after interruption

Read `marker.txt` from the current workspace, reverse its complete non-empty line, and return the decoded value in the required marker format. This is a narrowly scoped helper for the DeepSeek Harness integration example.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Decoding the marker.txt fixture
• Checking a DeepSeek Harness integration example
• Returning the decoded marker in a fixed format

Inspect a codebase to see its file count, total lines and size, file types, and largest files. The report groups files by extension and sorts those summaries by line count, giving you a quick view of project composition.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Checking how many lines of code a project contains
• Seeing which file types make up a codebase
• Finding the largest source files

Use this skill to break larger work into manageable turns, save checkpoints before time runs out, and make review requests and verdicts traceable to a specific change. It also guides cleanup of temporary checkouts.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Splitting a large task into reviewable steps
• Saving progress before an agent run times out
• Requesting and recording a review verdict

Prepare release documentation for Struts 6.x and 7.x releases or release candidates. It helps derive version notes from JIRA and merged changes, update the Migration Guide index, organize GitHub release notes, and draft a test-build announcement.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Preparing Struts version notes from JIRA and merged changes
• Updating the Migration Guide release index
• Organizing GitHub release notes by change type

Draft the formal [VOTE] email for an Apache Struts release candidate after its version notes, GitHub release, and staged artifacts are available. The skill renders the published notes into a fixed mail template and prepares a draft for review.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Drafting a Struts release candidate vote email
• Turning published version notes into the vote message
• Preparing a release vote draft for a release manager

Review Apache Struts pull requests for repository-specific test conventions, security-sensitive framework changes, implementation patterns, and commit hygiene. It helps distinguish blocking correctness or security concerns from minor suggestions.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing tests in a Struts pull request
• Checking changes to OGNL or file upload handling
• Spotting tests that pass without exercising behavior

Triage Dependabot pull requests by checking the dependency’s actual Maven scope and required build results. The skill identifies which updates need a Jira ticket, organizes the approval checkpoint, and guides the correct PR title, body, and squash commit details before landing.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Sorting an open Dependabot PR queue
• Checking whether a dependency update needs a Jira ticket
• Verifying required build checks before merging

Use the `gh` command-line tool to read and update GitHub issues, pull requests, repositories, releases, and workflow runs. Fetch diffs, inspect CI checks and logs, search code, or create issues and comments with structured output.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Checking a pull request diff and CI status
• Creating an issue or commenting on a PR
• Listing recent workflow runs and reading logs

Write, edit, or review C source and compatible header files using a project’s coding rules. Apply semantic rules while coding, then use clang-format to format or check the changed files.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Starting a C file that must follow a project style guide
• Reviewing .c or .h code for naming and documentation rules
• Formatting changed C files with clang-format

A small fixture demonstrating a YAML folded description that can parse with a trailing newline. It provides an example for recognizing why description text may need normalization before being placed in Markdown tables.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Checking how a folded YAML description parses
• Reproducing a trailing newline in parsed text
• Considering normalization before Markdown table output

Handle Git commits, branches, merges, conflicts, worktrees, and releases with clear change tracking. The skill supports Conventional Commits, explicit file staging, and repository governance checks such as ownership, CI expectations, and audit evidence.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Preparing a safe, well-scoped Git commit
• Resolving merge conflicts or planning a rebase
• Creating branches, tags, or worktrees

Create or revise agent skills in the SSL-lite Markdown format, with clear activation rules, workflow, resources, and safety details. The skill can also convert an existing skill or review whether its instructions are complete enough to use and validate.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Creating a new agent skill
• Converting an existing skill to SSL-lite
• Reviewing skill routing and execution steps