Source profileQuality 86/100

ros-claw/rosclaw/.agents/skills/rosclaw/SKILL.md

rosclaw

Use when operating, validating, or changing ROSClaw physical-AI runtime workflows, especially CLI smoke tests, Practice evidence loops, body/runtime checks, MCP integration, MuJoCo sandbox verification, and safe ROS or hardware boundaries.

Source repository stars
178
Declared platforms
0
Static risk flags
0
Last source update
2026-08-04
Source checked
2026-08-04
DecisionCompatibilityInstallWorkflowPermissionsEvidenceSource

Decision brief

What it does—and where it fits

Use this exact launcher for every ROSClaw CLI command: .venv/bin/python -m rosclaw.entrypoint. Do not replace it with a different rosclaw found on PATH.

Best for

  • Use when operating, validating, or changing ROSClaw physical-AI runtime workflows, especially CLI smoke tests, Practice evidence loops, body/runtime checks, MCP integration, MuJoCo sandbox verification, and safe ROS or…

Not for

  • Tasks that require unconfirmed production actions or broad system permissions.
  • Environments where the pinned source and install steps cannot be inspected.

Compatibility matrix

Platform support, with evidence labels

PlatformStatusEvidenceWhat to check
CodexNot declaredNo explicit evidencePortability before use
Claude CodeNot declaredNo explicit evidencePortability before use
CursorNot declaredNo explicit evidencePortability before use
Gemini CLINot declaredNo explicit evidencePortability before use
Open the compatibility checker

Installation

Inspect first. Install second.

The source command is displayed only when detected. A safe inspection prompt is always available so your agent can explain every action before execution.

Source-detected install commandSource
npx skills add https://github.com/ros-claw/rosclaw --skill ".agents/skills/rosclaw"
Safe inspection promptEditorial

Inspect the Agent Skill "rosclaw" from https://github.com/ros-claw/rosclaw/blob/46f7167880e9248a13149bd8a497905d9c19ac72/.agents/skills/rosclaw/SKILL.md at commit 46f7167880e9248a13149bd8a497905d9c19ac72. List every install step, command, network request, credential, file read/write, external action, and rollback step. Explain whether it fits my task. Do not install or execute anything until I approve.

Workflow

What the source asks the agent to do

  1. 01

    Robot Integration Workflow

    Robot Integration installation/configuration is a CLI lifecycle around the MCP runtime. It is not an additional MCP tool surface.

    robot install verifies and installs the signed contract. PassBind a discovered device with robot configure only when an exact model andA read-only hardware check requires a configured instance and canonical
  2. 02

    Capability App Workflow

    Apps call named Capabilities only. They do not install drivers, issue

    Apps call named Capabilities only. They do not install drivers, issueRun SHADOW first with app run --body --mode SHADOW --json.- Apps call named Capabilities only. They do not install drivers, issue permits, arm rosclawd, or authorize direct hardware access. - Run SHADOW first with app run --body --mode SHADOW --json. Treat a local success as c…
  3. 03

    LeRobot Bridge Workflow

    These are operator workflows. Do not install dependencies unless explicitly requested.

    getproductstatusgetruntimestatusgetbodyprofile
  4. 04

    Operator setup

    These are operator workflows. Do not install dependencies unless explicitly requested.

    These are operator workflows. Do not install dependencies unless explicitly requested.
  5. 05

    Dataset workflow

    For non-physical data operations, use the ROSClaw CLI:

    For non-physical data operations, use the ROSClaw CLI:

Permission review

Static risk signals and limitations

No configured static risk pattern was detected

This is not proof of safety. Runtime behavior, indirect dependencies, and hidden external systems are outside the static scan.

Evidence record

Why each signal appears

EvidenceSourceComputedTestedEditorial
SignalValueEvidence typeMeaning
Quality score86/100ComputedDocumentation, specificity, maintenance, and trust rules
Repository stars178SourceRepository attention, not individual Skill quality
Compatibility0 platformsSourceDeclared in the catalog source record
Usage guideautomated source guideEditorialGenerated or reviewed according to the visible evidence level

Pinned source

Provenance and original SKILL.md

Repository
ros-claw/rosclaw
Skill path
.agents/skills/rosclaw/SKILL.md
Commit
46f7167880e9248a13149bd8a497905d9c19ac72
License
MIT
Collected
2026-08-04
Default branch
main
View the original SKILL.md

ROSClaw Agent Skill

Use this exact launcher for every ROSClaw CLI command: .venv/bin/python -m rosclaw.entrypoint. Do not replace it with a different rosclaw found on PATH.

Safety

  • Treat ROSClaw as physical-AI infrastructure. Do not publish ROS topics, actuate hardware, run real robot skills, or mutate a live workspace directly, even when a user asks for that specific action.
  • Prefer dry-run, read-only, mock, fixture, simulation, or temp-workspace commands for validation.
  • Use a temporary ROSCLAW_HOME for CLI smoke tests that write persistent state.
  • Prefer --json for machine checks, then validate the JSON parses.
  • The only Agent-side action entry is MCP request_action, backed by rosclawd. Never instantiate a local Runtime or register a hardware driver/executor from the Agent process.
  • REAL actions require an immutable body snapshot, a daemon-issued capability- and action-intent-bound permit, and a registered verified REAL executor. Refuse direct ROS, DDS, serial, CAN, SDK, or motor commands.

First Checks

.venv/bin/python -m rosclaw.entrypoint doctor --json
.venv/bin/python -m rosclaw.entrypoint status capabilities
.venv/bin/python -m rosclaw.entrypoint daemon status --json
.venv/bin/python -m rosclaw.entrypoint demo list
.venv/bin/python -m rosclaw.entrypoint demo run ur5e-reach
.venv/bin/python -m rosclaw.entrypoint explain latest
.venv/bin/python -m rosclaw.entrypoint agent doctor universal --project-root .
.venv/bin/python -m rosclaw.entrypoint agent test universal --project-root . --quick --mcp-probe

For daemon status, require running and ledger.integrity_verified to be true; require ledger.write_failed, recovery.required, and emergency_stop_latched to be false. Production REAL work also requires supervision_state=ARMED, privilege_separated=true, and .venv/bin/python -m rosclaw.entrypoint daemon security-check --json with boundary_ready=true, daemon_uid_pinned=true, and ledger_state_private=true; same-UID development proves only process separation. Read an existing durable result with .venv/bin/python -m rosclaw.entrypoint daemon action-status <ACTION_ID> --json and .venv/bin/python -m rosclaw.entrypoint daemon receipt <ACTION_ID> --json. Do not use daemon acknowledge-recovery as Agent automation: it is an operator incident-review command and does not clear E-stop or prove physical state.

Robot Integration Workflow

Robot Integration installation/configuration is a CLI lifecycle around the MCP runtime. It is not an additional MCP tool surface.

.venv/bin/python -m rosclaw.entrypoint robot discover --json
.venv/bin/python -m rosclaw.entrypoint robot install realsense --json
.venv/bin/python -m rosclaw.entrypoint robot verify realsense --stage contract --json
  • robot install verifies and installs the signed contract. Pass --install-adapter only when the operator explicitly authorizes dependency installation in the selected Python environment.
  • Bind a discovered device with robot configure only when an exact model and stable serial are available. --allow-offline creates configuration, never hardware evidence.
  • A read-only hardware check requires a configured instance and canonical rosclawd receipt. Treat local success as candidate evidence; never promote product support or claim physical success from CLI text alone.
  • Do not call the vendor SDK directly. Runtime device access stays behind MCP request_action and the daemon boundary.

Capability App Workflow

.venv/bin/python -m rosclaw.entrypoint app list
.venv/bin/python -m rosclaw.entrypoint app validate realsense-inspect --json
  • Apps call named Capabilities only. They do not install drivers, issue permits, arm rosclawd, or authorize direct hardware access.
  • Run SHADOW first with app run <APP> --body <BODY> --mode SHADOW --json. Treat a local success as component evidence, never an independent hardware or Agent verification claim.

Practice Evidence Loop

Use fixture-based Practice workflows before touching real robots:

TMP=$(mktemp -d /tmp/rosclaw-practice.XXXXXX)
export ROSCLAW_HOME="$TMP/home"
.venv/bin/python -m rosclaw.entrypoint practice record --fixture tests/fixtures/practice/rh56_minimal_loop.json --out "$TMP/practice" --json
.venv/bin/python -m rosclaw.entrypoint practice verify practice_rh56_minimal_loop --data-root "$TMP/practice" --strict --json
.venv/bin/python -m rosclaw.entrypoint practice distill practice_rh56_minimal_loop --data-root "$TMP/practice" --json

MCP Contract

  • Expected P0 tools: get_robot_state, list_skills, query_memory, validate_trajectory, sandbox_run, practice_query, emergency_stop, get_body_profile, get_body_state, list_body_capabilities, query_body, validate_body_action, get_calibration_status, get_runtime_status, request_action, get_action_status, cancel_action, get_product_status, list_product_demos, run_product_demo, get_execution_receipt, explain_execution.
  • run_product_demo executes an official MuJoCo path and persists an integrity-checked ExecutionReceipt; it never commands real hardware.
  • Use get_execution_receipt and explain_execution instead of inferring success from text output.
  • sandbox_run is simulation-only. If no physics state is available, treat the result as degraded rather than live.
  • validate_trajectory can approve a plan, but it does not authorize direct hardware execution.
  • request_action sends an unapproved envelope to rosclawd; only the daemon may turn a matching server-issued permit into authorization.
  • cancel_action never claims that active physical motion stopped. Use emergency_stop and the certified hardware E-stop when motion may be active.

LeRobot Bridge Workflow

Operator setup

These are operator workflows. Do not install dependencies unless explicitly requested.

.venv/bin/python -m rosclaw.entrypoint setup lerobot --reference-policy rh56
.venv/bin/python -m rosclaw.entrypoint lerobot doctor --json

Read-only Agent discovery

Use MCP tools:

  1. get_product_status
  2. get_runtime_status
  3. get_body_profile
  4. get_body_state
  5. get_calibration_status
  6. list_body_capabilities

SHADOW request

Submit through request_action with execution mode SHADOW. Poll with get_action_status. Read the final result with get_execution_receipt. Explain it with explain_execution.

REAL request

Submit the exact requested action through request_action. The Agent does not issue or approve a Permit. If rosclawd reports AUTHORIZATION_REQUIRED, stop and explain the missing authorization. Never fall back to shell, serial, CAN, ROS topic, MCP driver, or vendor SDK execution.

Dataset workflow

For non-physical data operations, use the ROSClaw CLI:

.venv/bin/python -m rosclaw.entrypoint practice verify <PRACTICE_ID> --strict --json
.venv/bin/python -m rosclaw.entrypoint practice export   --practice-id <PRACTICE_ID>   --format lerobot   --profile physical   --output <OUTPUT_DIR>   --json

Unsupported requests

Training, DAgger, reward models, Hub publishing, arbitrary-policy robot mapping, CAN RH56 execution, open-loop chunks, and unattended execution are outside LeRobot Bridge v1.0.1.

Decision table

User intentAgent should doAgent must not do
"Is LeRobot installed?"get_product_status + get_runtime_statusimport lerobot
"Can I use the RH56 now?"body/profile/state/calibration/capability toolsopen the serial port
"Preview what it would do"request_action(SHADOW)run rollout execute directly
"Do the real OK gesture"request_action(REAL), wait for rosclawdself-issue a Permit
"Try it without authorization"refuse and explain AUTHORIZATION_REQUIREDswitch to serial commands
"Did it succeed?"get_action_status + receipt + explain_executionguess success from text
"Export data for training"practice verify / exporttrain automatically
"Control RH56 with any VLA"explain v1.0.1 supports only the reference contractguess mapping by action_dim
"Control RH56 over CAN"explain CAN is fail-closed unsupportedreuse RS485 mapping on CAN

Alternatives

Compare before choosing

Computed 10023,781

alirezarezvani/claude-skills

app-store-optimization

App Store Optimization (ASO) toolkit for researching keywords, analyzing competitor rankings, generating metadata suggestions, and improving app visibility on Apple App Store and Google Play Store. Use when the user asks about ASO, app store rankings, app metadata, app titles and descriptions, app store listings, app visibility, or mobile app marketing on iOS or Android. Supports keyword research and scoring, competitor keyword analysis, metadata optimization, A/B test planning, launch checklist

2026-07-170 platforms
Computed 10014,225

wanshuiyin/Auto-claude-code-research-in-sleep

citation-audit

Use it for operations and research tasks; the detail page covers purpose, installation, and practical steps.

2026-08-040 platforms
Computed 1004,922

dotnet/skills

migrate-vstest-to-mtp

Migrates .NET test projects from VSTest to Microsoft.Testing.Platform (MTP). Use when user asks to "migrate to MTP", "switch from VSTest", "enable Microsoft.Testing.Platform", "use MTP runner", set OutputType=Exe only for test projects in Directory.Build.props, or mentions EnableMSTestRunner, EnableNUnitRunner, or UseMicrosoftTestingPlatformRunner. USE FOR: MTP behavioral differences vs VSTest (exit code 8, zero tests discovered, --ignore-exit-code, TESTINGPLATFORM_EXITCODE_IGNORE); centralizing

2026-08-040 platforms
Computed 1002,504

aaron-he-zhu/aaron-marketing-skills

social-selling-planner

Use when the user asks to "set up my founder social-selling routine", "build a daily engagement block for target accounts", or "turn funding / hiring signals into selling plays"; produces the founder/seller daily operating block — a time-boxed engagement-block spec (substantive value-add comments on target-account posts, never a pitch), warm-touch-before-ask cadence rules, trigger-response plays consuming the social-pulse-monitor B2B trigger watchlist (funding / hiring / launch signals), and a q

2026-08-041 platforms