AgentStack
SKILL verified MIT Self-run

Household Open Task

skill-miaodx-roboclaws-household-open-task · by MiaoDX

Complete open-ended household robot goals through public household MCP tools.

No reviews yet
0 installs
10 views
0.0% view→install

Install

$ agentstack add skill-miaodx-roboclaws-household-open-task

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Household Open Task? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Household Open Task

Use the public household MCP tools as a bounded robot capability surface. The operator goal is authoritative. Do not start a room-cleanup routine unless the operator explicitly asks for cleanup.

Loop

  1. Call roboclaws__metric_map() when map context is needed. Use Base

Metric Map waypoints, public room labels, Runtime Metric Map evidence, and roboclaws__resolve_target_query() for named places, stale labels, or search terms. Do not read private manifests, scene inventory, generated mess truth, or scoring artifacts.

  1. Inspect only as much as the goal needs. Navigate to public waypoints or

target candidates with roboclaws__navigate_to_waypoint(), then call roboclaws__observe(). Use roboclaws__adjust_camera() only for bounded public recovery when a target or observation is incomplete.

  1. For information, search, or inspection goals, tie the final answer to public

observations, target candidates, and the inspected search budget. A not-found answer is valid only after public evidence shows the useful search space has been checked or exhausted.

  1. For manipulation goals, act only on task-relevant observed handles or visual

candidates. Follow public tool recovery responses such as required_tool, required_next_tool, blocked_capability, and actionability status. If the backend blocks manipulation, report the blocker and call done with the public evidence gathered so far.

  1. Call roboclaws__done(reason) when the operator goal is satisfied, blocked

by a public capability response, or exhausted by the public search budget. The reason must summarize the public evidence used and any remaining risk.

Boundaries

  • Do not call scene_objects, inspect private evaluator files, infer hidden

mess targets, or treat cleanup candidates as part of the goal unless the operator asked for cleanup.

  • Do not require every inspection waypoint to be visited unless the goal itself

asks for a full-room sweep or a preset selects that policy.

  • Do not promote this into one opaque MCP task tool. Keep search, navigation,

observation, manipulation, and done visible in the trace.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.