菜单
Manages the lifecycle of Cuttlefish instances and their groups using podcvd, including automatic management of all ADB connections.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/google/android-cuttlefish --skill podcvd命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | podcvd |
| description | Manages the lifecycle of Cuttlefish instances and their groups using podcvd, including automatic management of all ADB connections. |
This skill fully orchestrates the lifecycle of Cuttlefish instances and their groups within container environments, including the automatic establishment and teardown of ADB connections.
[!WARNING] Exclusive ADB Connection Control: This skill internally manages all required ADB connections when creating, removing, or clearing instances. You MUST NOT invoke external skills, MCP servers, or secondary tools to establish ADB connections to these Cuttlefish instances, as doing so will cause system conflicts.
[!WARNING] Strict Isolation from
cvd:podcvdand standardcvddo not share state or information. Mixing their usage will cause inconsistent states and system errors.
- MUST NOT use standard
cvdcommands if thispodcvdskill has been used in the environment.- MUST NOT use this
podcvdskill if standardcvdwas previously used to manage instances.- Why Instances May Seem Missing: Because they do not share state, instances created via standard
cvdwill not be visible topodcvd(e.g.,podcvd fleetwill show no instances), and vice versa. If you cannot find an expected Cuttlefish instance, verify which tool was used to create it.
lunch <target>) and build the required targets (e.g., m) to ensure that valid Cuttlefish host tools (inside ANDROID_HOST_OUT) and Android device images (inside ANDROID_PRODUCT_OUT) are generated.ANDROID_HOST_OUT and ANDROID_PRODUCT_OUT).podcvd Command: If podcvd is not installed (e.g., which podcvd fails), install the package using sudo apt install cuttlefish-podcvd -y.podcvd fails due to an unconfigured host (e.g., explicit prompts to run podcvd-setup or missing device permissions such as /dev/kvm), execute podcvd-setup to initialize the host.Setting the PODCVD_CLIENT_ID environment variable is MANDATORY whenever executing podcvd subcommands. This ensures proper session tracking and prevents resource collisions.
references/podcvd_client_id.md exists within this skill directory.references/podcvd_client_id.md exists: You MUST read its contents using view_file and strictly follow the platform-specific instructions provided inside to construct and pass PODCVD_CLIENT_ID before calling podcvd.references/podcvd_client_id.md does NOT exist: Only as an exception, omit setting PODCVD_CLIENT_ID (leave it empty) and execute podcvd directly.[!IMPORTANT] Always follow these essential execution policies when managing Cuttlefish instances:
create, start, restart, powerwash) automatically establish or re-establish ADB connections, while subcommands (e.g., stop, remove, clear, reset) automatically tear down ADB connections. Never invoke external ADB tools manually.podcvd create initializes a new instance group and returns a JSON response containing its assigned "group_name". Because "group_name" is the primary identifier for managing that group, you MUST extract and track this string (from the create output or podcvd fleet) to dynamically pass --group_name=<group_name> in subsequent lifecycle operations (e.g., stop, start, restart, remove).create, start, restart, powerwash) takes approximately 1 minute for the virtual device boot process and ADB connection setup to complete. During this boot window, podcvd fleet will report status "Starting" and adb devices -l may temporarily show an empty list. You MUST wait for podcvd fleet status to transition to "Running" (up to 1-2 minutes) before verifying ADB device readiness or executing teardown operations. Do not prematurely cancel, clear, or report failure while status is still "Starting".podcvd fleet BEFORE and AFTER state transitions. Before issuing any state-changing subcommand (e.g., stop, restart, remove), execute podcvd fleet to ensure previous operations have finished and the group is in the expected state (e.g., "Running"). After transitions, verify status and confirm ADB connection readiness using podcvd fleet and adb devices -l.To ensure high success rates and avoid malformed options, prefer using the fully constructed standard command lines below for general tasks. Always include PODCVD_CLIENT_ID="<client_id>" at the start of every command line.
Creates an instance group with standard configurations and automatically establishes ADB connections. (For bulk creation, see the concurrency policy below).
PODCVD_CLIENT_ID="<client_id>" podcvd create --vhost_user_vsock=true --report_anonymous_usage_stats=n
Lists all currently running Cuttlefish instance groups.
PODCVD_CLIENT_ID="<client_id>" podcvd fleet
Removes a targeted instance group and automatically destroys corresponding ADB connections. Replace <group_name> with the actual ID.
PODCVD_CLIENT_ID="<client_id>" podcvd --group_name=<group_name> remove
Completely tears down all Cuttlefish instance groups and destroys all relevant ADB connections globally.
PODCVD_CLIENT_ID="<client_id>" podcvd clear
Unless the user explicitly requests sequential execution, operations targeting different instance groups (such as creating multiple instances or removing multiple distinct groups) MUST ALWAYS be executed in parallel.
User intent is paramount and overrides this rule, but parallel processing is the mandatory default for any bulk workload.
[!WARNING] Default Strict Prohibition: When handling bulk requests (e.g., "create 10 instances" or "remove groups A, B, and C"), you MUST NOT execute them sequentially by default. Waiting for one task to complete before initiating the next severely degrades system performance.
Execution Guidance for Agents:
Launch independent podcvd commands concurrently as background tasks, with leveraging agent's native capability dealing with concurrent behavior if possible.
Exceptions to Parallel Execution:
group_name (e.g., creating a group and immediately modifying it).clear subcommand internally processes global cleanup across all groups, so manual parallelization does not apply.Below are the currently supported podcvd subcommands in alphabetical order:
bugreport: Captures diagnostic host logs into a zip archive for troubleshooting (optionally includes device adb bugreport)cache: Manage the files cached by cvdclear: Clears the instance database, stopping any running instances firstcreate: Create a Cuttlefish instance groupdisplay: Enables hotplug/unplug of displays from running cuttlefish virtual devicesenv: Enumerate + Query APIs for all gRPC services made available by this virtual device instancefetch: Retrieve build artifacts based on branch and target namesfleet: Lists active devices with relevant informationhelp: Used to display help information for other commandslint: Error checks the input virtual device json config fileload: Creates and starts an instance group from a JSON configuration filelogin: Acquire credentialslogs: List and display log filesmonitor: Monitor device logs (launcher, kernel, and logcat) in real-timepowerbtn: Trigger power button event on the devicepowerwash: Reset device to first boot stateremove: Remove instance groups and related artifacts from the systemreset: Remove all instance groups and kills orphaned cuttlefish processesrestart: Restart deviceresume: Resume the cuttlefish devicescreen_recording: Record screen contentssetup: Configure the host for Cuttlefishsnapshot_take: Take snapshot of the cuttlefish devicestart: Start all Cuttlefish Instances in a groupstatus: Query status of a single instance group. Use podcvd fleet for all devicesstop: Stop Cuttlefish instancessuspend: Suspend the cuttlefish deviceversion: Prints version of cvd client and cvd serverTo inspect supported flags and advanced setup options for any subcommand, query built-in help directly:
PODCVD_CLIENT_ID="<client_id>" podcvd help <subcommand>
Executing the create or fleet subcommands returns JSON outputs containing instance configuration details. When reporting these connection details in your responses, you MUST follow these exact structural formats:
IP:PORT format (for example: 192.168.112.1:6520). Never output only the port number or adb_port: 6520.web_access URL and present ONLY the base Protocol, IP, and Port (for example, if JSON returns "https://192.168.112.1:11443/devices/cvd_1-1-1/files/client.html", present it as https://192.168.112.1:11443).