stateport-capture
Capture a browser state or reproduce supplied steps with StatePort, including login using supplied credentials, and review and save the result as a reusable State Card.
StatePort-Dev · 已有静态检查记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add StatePort-Dev/stateport-agent-skills --skill stateport-capture技能说明
name: stateport-capture description: Capture a browser state or reproduce supplied steps with StatePort, including login using supplied credentials, and review and save the result as a reusable State Card. license: MIT
StatePort Capture
Create the state the user asked to record. It may be a bug reproduction, an ordinary application state, or a reusable interaction sequence. Follow the Capture procedure, workflow guidance and data handling.
Use the URL, steps, login data, desired outcome and display preferences already in the task. Choose a descriptive Card name if none was given. Keep the runtime viewport and headless defaults unless the task needs another size or the user wants to watch. Ask only for information needed to perform the intended task, such as an absent URL or a genuinely ambiguous action.
When the task includes bug information, including a link to an issue, use that information to populate Bug context: Description, Observed behavior and Expected behavior. Read a supplied issue through an available authorized connector or browser if its contents are not already present. Do not ask the user to repeat accessible information or confirm copying it. Follow the Bug context procedure to map the content, distinguish reported symptoms from this attempt's observations, and save it using the connected runtime's supported fields. If persistence is unavailable, provide the prepared fields and state that they were not saved.
Discover the public schemas and call get_capture_capability. With Capture
available and enabled, start recording before following the steps. Use the
exact managed Page through observe_capture_page and act_on_capture_page.
Prefer fresh elementId targets and supported actions; observe again after
each action. Follow supplied steps in order, including multi-page login.
Use login credentials supplied for this task through start_capture.credentials
and fill_credential when advertised. There is no extra consent question for
the requested login, and no ticket reference is needed for this path. Never
repeat the values in commentary, reports or ordinary fill arguments. Older
runtime alternatives and recovery are described in the Capture procedure.
When the requested state is reached, stop into Review, inspect the result and save the requested Card. A request to capture and save covers this sequence; Review is a runtime step, not a reason to ask again. Confirm a bug symptom only when the user asked to reproduce that symptom. An inconclusive attempt may be saved if the user wants it retained, with its actual status clearly described.
Return the saved Card/revision and a concise account of the state and steps captured, plus any gaps. For a failed attempt, use supported recovery or discard as appropriate and explain the actionable blocker. Continue into debugging, Journey or transfer only when the user's task calls for it; the corresponding procedures are included in this skill's references.