开发技能

Use this skill to get oriented in a code repository before making changes. It guides the agent to review architecture notes, find the relevant module, and identify suitable verification commands for the work.

中文简介正在准备中。当前显示英语简介。

• Understanding a repository’s architecture
• Finding the module responsible for a feature
• Choosing files to inspect before coding

Trace a Claude Code session through Agent Monitor data, including its event timeline, tool calls, agent tree, token usage, and cost. Use a session ID, the latest session, or recent errors to produce a report with anomalies and a likely diagnosis.

中文简介正在准备中。当前显示英语简介。

• Investigating why a session failed
• Tracing tool calls and agent activity
• Reviewing token use and session timing

Inventory configured MCP servers across user and project settings. Review each server’s transport, launch command or URL, environment variable or header names, and source file to spot duplicates, unknown definitions, or remote endpoints that need review.

中文简介正在准备中。当前显示英语简介。

• Checking which MCP servers are configured
• Finding duplicate servers across scopes or files
• Reviewing stdio commands and environment variable names

Add a file overview and the required author line to applicable source files when creating or editing them. Check repository-wide header compliance with the bundled audit scripts across supported file types.

中文简介正在准备中。当前显示英语简介。

• Adding headers to newly created source files
• Updating a touched file that lacks a header
• Checking header compliance across a repository

Investigate production-like failures by recording symptoms, reproducing them with a small test case, and checking logs, API responses, or database state. Isolate the affected subsystem, confirm the root cause, then make and verify a minimal fix.

中文简介正在准备中。当前显示英语简介。

• Tracing a regression across API, database, and UI behavior
• Investigating flaky workflows with reproducible steps
• Checking data inconsistencies using logs and database state

Search a Claude Code session transcript for a word or regular expression, then review each matching message with nearby context. The results include relevant roles, timestamps, tool details, and a summary of how many messages matched.

中文简介正在准备中。当前显示英语简介。

• Find a prompt, error, or tool call in the latest session
• Locate messages that contain a particular phrase
• Review matches for a regular expression across a session transcript

Audit a Claude Code setup through the Agent Monitor Config Explorer API. Inventory user and project settings, find duplicate skills or agents, flag hooks that run shell commands, and distinguish surfaces the explorer can edit from those that require manual changes.

中文简介正在准备中。当前显示英语简介。

• Reviewing a Claude Code setup for duplicate skills
• Checking hooks that execute shell commands
• Understanding which configuration surfaces are editable

Review a single Claude Code session as a chronological timeline of events, tool names, and summaries. The report highlights long pauses, missing tool completions, failures, compactions, and missing session boundaries to help explain what happened step by step.

中文简介正在准备中。当前显示英语简介。

• Tracing what happened during a Claude Code session
• Finding tool calls with no recorded completion
• Checking a session for errors or long pauses

Keep localized content aligned across dashboard text, wiki pages, mirrored guides, language switchers, and locale-aware formatting. Use this workflow when changing user-facing copy, translation keys, guides, wiki content, or adding a supported language, with English as the source for each surface.

中文简介正在准备中。当前显示英语简介。

• Add or change a dashboard translation key
• Update English guide content across translated guides
• Add a supported language across the product surfaces

Set up and launch CCAM’s MCP server by installing dependencies, building it, and choosing stdio, HTTP, or REPL transport. Configure mutation and destructive-operation safeguards, dashboard authentication, then validate that the tool catalog is registered consistently.

中文简介正在准备中。当前显示英语简介。

• Building the MCP server after installing dependencies
• Choosing stdio, HTTP, or REPL transport
• Setting write and destructive-operation safeguards

Guide medium or large feature changes through repository exploration, a short plan, focused implementation, relevant verification, and documentation updates. Use this workflow to keep changes scoped and preserve API and websocket contracts unless a change is intentional.

中文简介正在准备中。当前显示英语简介。

• Planning a feature change across repository modules
• Implementing a focused backend, frontend, or MCP change
• Running relevant checks after a feature implementation

Use this skill to update an existing GitHub pull request when its branch lives in a fork. It checks the PR and push permissions, verifies the local commit relationship, and guides explicit staging, verification, and pushing to the fork branch.

中文简介正在准备中。当前显示英语简介。

• Updating an existing pull request from a fork
• Checking whether maintainers can push to a PR branch
• Verifying changes before pushing to a fork

Prepare one identified pull request by addressing review feedback, updating it with the latest base branch, and resolving conflicts. Check reviews and required status checks, then merge only after the candidate commit is confirmed ready.

中文简介正在准备中。当前显示英语简介。

• Address review feedback and merge a pull request
• Resolve conflicts with the base branch before merging
• Check whether a pull request meets its merge requirements

Preview completed Git branches and worktrees that are safe to remove, then clean them up locally. With an additional option, delete matching remote branches on GitHub. The workflow skips protected branches, open pull requests, and worktrees with uncommitted changes.

中文简介正在准备中。当前显示英语简介。

• Removing local branches after pull requests merge
• Tidying completed worktrees safely
• Deleting merged GitHub branches when requested

Find local Git worktree folders whose branches or pull requests have merged into main, then preview which clean, inactive folders can be removed. The bundled script defaults to a dry run and requires an explicit apply option to perform cleanup.

中文简介正在准备中。当前显示英语简介。

• Previewing merged worktree folders for cleanup
• Tidying old feature branch directories
• Removing merged PR worktrees while keeping branch refs

Flash or update EspControl display firmware with ESPHome. Choose a supported display or the full device set, then upload through an explicitly provided OTA target or USB connection using the repository’s device configurations.

中文简介正在准备中。当前显示英语简介。

• Updating firmware on a supported display
• Reflashing all known displays in sequence
• Uploading firmware to a display over USB

Inspect unresolved review feedback across all open pull requests authored by the user. Make focused fixes on each PR’s branch, run relevant checks, push changes, and resolve only the review threads those changes address.

中文简介正在准备中。当前显示英语简介。

• Inspect unresolved review comments across all your open pull requests
• Fix actionable feedback separately on each pull request branch
• Push verified updates and resolve the review threads they address

Audit static text shown on the EspControl device display and connect it to the firmware translation workflow. It helps identify missing translation keys, untranslated display strings, and language file entries, then regenerate and check the firmware translations.

中文简介正在准备中。当前显示英语简介。

• Adding a missing translation for a device screen label
• Auditing hard-coded text shown on the physical display
• Fixing firmware code that displays raw English text

Review a pull request for bugs, suspicious code, dependency and workflow risks, and project-specific issues. Get a practical merge recommendation, focused check results, and clear findings before deciding whether to merge.

中文简介正在准备中。当前显示英语简介。

• Assess whether an open pull request is safe to merge
• Inspect an outside contribution for suspicious changes
• Find bugs or missing checks in a proposed change

Compile ESPHome firmware for every listed device using the version pinned by the repository and Docker. When a build fails, inspect the reported YAML or configuration issue, fix its source, and retry until each device compiles.

中文简介正在准备中。当前显示英语简介。

• Checking firmware builds before committing
• Finding YAML or configuration errors
• Compiling all listed devices with Docker

Review a project as if rebuilding it today while keeping its current user experience as the target. Produce evidence-based architecture and migration recommendations, explain tradeoffs in plain language, and keep the work research-only unless implementation is requested separately.

中文简介正在准备中。当前显示英语简介。

• Exploring a from-scratch architecture for a project
• Comparing current structure with a reset design
• Planning a staged migration without changing code

Review a codebase for maintainability problems such as duplicated logic, fragile assumptions, dead code, and confusing structure. Produce a prioritized cleanup backlog with file locations, practical risks, and suggested first steps; the scan itself does not implement fixes.

中文简介正在准备中。当前显示英语简介。

• Find code that may become difficult to maintain
• Build a prioritized technical debt cleanup backlog
• Review duplicated logic, dead code, or fragile assumptions

Address actionable feedback on an open pull request by reviewing unresolved comments, making focused code or documentation changes, and running relevant checks. After committing and pushing the update, resolve only the review threads that the changes actually address.

中文简介正在准备中。当前显示英语简介。

• Implement requested changes from unresolved PR comments
• Update documentation in response to review feedback
• Push fixes and resolve addressed review threads

Use this skill to resolve merge conflicts in an open pull request by bringing in its base branch and preserving the PR’s intended changes. It guides conflict review, focused checks, and branch updates while leaving the pull request open for testing.

中文简介正在准备中。当前显示英语简介。

• Update a PR branch with the latest base branch
• Resolve conflicts that prevent a PR from merging
• Make a conflicted pull request ready for testing