@hybridlabor-api/aos 4.13.2 → 4.14.0

package/skills/global_config/stay-within-limits/SKILL.md

@@ -0,0 +1,85 @@
+---
+name: stay-within-limits
+description: Use when long-running or parallel agent work must respect 5-hour and weekly usage limits by checking usage between waves, pausing near the cap, and resuming only when the window is clear.
+category: bdb-core
+source: BuilderIO/skills
+---
+
+# Stay Within Limits
+
+Keep long-running agent work inside the current 5-hour and weekly usage windows.
+Check usage before launching substantial work and between waves of parallel
+subagents. If an active 5-hour or weekly limit is at or above 95%, pause new
+work until the window is clear enough to continue safely.
+
+## Core Loop
+
+1. Run a bounded wave of work. Default to at most 3 parallel subagents unless
+   the user or host gives a different throttle.
+2. Wait for the wave to finish. Do not interrupt in-flight subagents just to
+   save budget; that usually loses work.
+3. Check current 5-hour and weekly usage with the host's usage/budget tool.
+4. If either window is at or above 95%, stop launching work and schedule a
+   self-contained resume when the relevant window should clear.
+5. On resume, re-check the real window or block before continuing. Do not trust
+   elapsed wall-clock time alone.
+
+## Usage Signals
+
+Prefer a first-party host usage tool when available. In Claude Code, use:
+
+```sh
+npx -y ccusage@20.0.26 blocks --active --json
+```
+
+Use the JSON to identify the active block start, current cost or percentage, and
+time remaining. On wake, compare the active block start timestamp with the
+previous one; a new timestamp is stronger evidence than "enough time passed."
+
+If the tool reports cost instead of a direct percentage, convert through the
+current account limit when known. For Claude Max-style 5-hour blocks, some users
+prefer an earlier caution threshold around $500-550; treat that as a
+user-configured guardrail, not a universal rule. The default stop rule is still
+95% of the active 5-hour or weekly limit.
+
+## Pausing And Resuming
+
+When a wake/resume tool is available, schedule a wakeup for:
+
+```txt
+min(3600, secondsUntilWindowClears)
+```
+
+If the runtime clamps wake delays to 60-3600 seconds, chain wakeups for longer
+waits. Each wakeup should re-check usage, reschedule if still over budget, and
+continue only when the window is safely below the threshold.
+
+Make wake prompts self-contained. Include:
+
+- The remaining plan.
+- The check-then-reschedule rule.
+- The 95% threshold and wave throttle.
+- The exact usage command or host usage tool to run.
+- The previous block/window identifier when available.
+- The next verification steps.
+- The next wave's handoff packets, including scope, verification commands, and
+  stop conditions, if delegation will resume.
+
+## Choosing The Wait Mechanism
+
+- Use a wake/resume tool when the agent needs instructions attached to the
+  future resume.
+- Use a background sleep or watcher for fixed timers and things a process can
+  observe directly.
+- Use cron or recurring schedules only for recurring fresh-session work.
+
+Avoid short-interval polling for things the host will notify you about, such as
+background task or subagent completion. For budget pauses, a prompt-cache miss
+after a long sleep is acceptable; preserving the limit matters more.
+
+## Reporting
+
+If you pause, tell the user which window is over threshold, the observed usage,
+when you scheduled or expect the next check, and what work remains. Keep enough
+state in the wake prompt that the next turn can resume without relying on
+conversation momentum.

package/skills/global_config/triage/SKILL.md

@@ -110,3 +110,6 @@ Capture everything resolved during grilling under "established so far" so the wo
 ## Resuming a previous session
 
 If prior triage notes exist on the issue or PR, read them, check whether the reporter has answered any outstanding questions, and present an updated picture before continuing. Don't re-ask resolved questions.
+
+## Visual recap
+To get oriented on a PR quickly, `/pr-recap --pr <n>` builds a local page from its diff (read-only `gh pr view`). It is informational and does not gate triage, and it never posts a comment.

package/skills/global_config/visual-edit/README.md

@@ -0,0 +1,96 @@
+# /visual-edit
+
+Open a running local app in the Agent-Native Design surface as URL-backed iframe
+screens for visual inspection, route-state exploration, and source-backed edits.
+
+Use `/visual-edit` when a UI needs to be reviewed or changed in context: compare
+real routes, inspect responsive states, walk a multi-screen flow, duplicate a
+screen for a new URL state, or apply visual changes back through the coding
+agent. The canvas uses the app's live routes and local bridge rather than a
+copied static HTML snapshot.
+
+## Install
+
+Run this from your app repository to install the skill and hosted Design
+connector:
+
+```sh
+npx @agent-native/core@latest skills add visual-edit
+```
+
+## Examples to try
+
+- **One screen:** `/visual-edit the dashboard at desktop size`
+- **Several screens:** `/visual-edit compare the dashboard, settings, and reports pages side by side`
+- **A full flow:** `/visual-edit the onboarding flow from welcome through account created`
+- **Responsive sizes:** `/visual-edit show the dashboard at desktop, tablet, and mobile widths`
+
+## Example: refine an onboarding flow
+
+Start your app locally, then ask your coding agent:
+
+```text
+/visual-edit the onboarding flow, plus home at every breakpoint
+```
+
+Design can place the four onboarding screens and Home at desktop, tablet, and
+mobile sizes together on one canvas. Select the **Get started** button, change
+its fill with the color picker, and drag its padding on the canvas to see the
+layout respond.
+
+When the result looks right, ask your coding agent to **Pull in my visual
+edits**. The agent retrieves the pending batch and updates the app's source for
+you to review. Design does not write those changes to your code automatically.
+
+## Choose the screens deliberately
+
+- **Multiple pages:** pass `paths` such as `["/", "/pricing", "/settings"]`.
+- **A multi-step flow:** pass named `routes` in the order the user follows,
+  such as shipping, payment, then confirmation. Each URL state becomes a
+  separate canvas screen.
+- **The same page at responsive widths:** add `viewports`, for example
+  `["desktop", "mobile"]`. Design places each requested route or page once per
+  viewport so they can be compared side by side.
+
+For example, add these fields to an `open-visual-edit` payload to place each
+checkout step at desktop and mobile sizes:
+
+```json
+{
+  "title": "Checkout flow review",
+  "routes": [
+    { "url": "http://localhost:5173/checkout?step=shipping", "title": "1. Shipping" },
+    { "url": "http://localhost:5173/checkout?step=payment", "title": "2. Payment" },
+    { "url": "http://localhost:5173/checkout?step=done", "title": "3. Confirmation" }
+  ],
+  "viewports": ["desktop", "mobile"]
+}
+```
+
+Pass that shape to `open-visual-edit` together with the connection details from
+the workflow. For a page comparison, use `paths` instead of `routes`.
+
+## Open the canvas beside the conversation when possible
+
+Use the Design URL or embed returned by `open-visual-edit`. If the current host
+supports an inline preview, webview, or side browser, open Design there so the
+canvas remains beside the chat. Host support varies; when no such surface is
+available, provide the returned **Open design** link instead.
+
+Prefer the interactive MCP App when the connected host renders it. In that
+surface, **Apply design updates** can submit the source-edit handoff to the
+current host conversation after its normal confirmation. An ordinary Browser
+pane has no trusted page-to-chat bridge: Apply uses Design's local agent there,
+and **Copy prompt to your agent** is the coding-agent fallback.
+
+Inside Design, choose **Show/Hide UI** from the `Cmd+K` menu or press Figma's
+`Shift+\` shortcut. The same action is available from Design's empty-canvas
+context menu.
+
+The hosted Design MCP connector handles the account-backed open, screen
+placement, and source-edit workflow. Public/read-only designs may be viewed
+without signing in; creating, saving, or sharing a design still requires an
+authenticated account.
+
+For the complete walkthrough and sharing details, see the [Visual Edit guide
+in Agent-Native](https://github.com/BuilderIO/agent-native/blob/main/skills/visual-edit/README.md).