agentsclimarketplace

Live testing

Skill Mistakey/simulink-automation-suite/.claude/skills/live-testing

Execute end-to-end functional testing on real MATLAB/Simulink hardware. Schema-driven discovery makes this version-agnostic — no updates needed when plugin actions change. Use when asked to "run live test", "test on real device", "实机测试", or "/live-test". Supports full, incremental, and targeted re-test modes.From its SKILL.md

Install
npx -y skills add Mistakey/simulink-automation-suite --skill live-testing

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 9 stars9 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.9 KB, as published. Nobody here has run it

Live Device Testing

End-to-end functional verification of the simulink-automation-suite plugin on real MATLAB/Simulink hardware. This skill tests the plugin as a user would — through its published interface, not its source code.

On-Demand References

  • .claude/skills/live-testing/phases.md — Detailed phase definitions, test case generation rules, pass/fail criteria, error handling. Read when starting test execution.
  • .claude/skills/live-testing/report-template.md — Report format, update rules, suggestions generation. Read when writing or updating the report (Phase 6).

Isolation Rules

PROHIBITED:
  - Read simulink_cli/ source code
  - Read tests/ unit tests

REQUIRED:
  - Discover available actions via: python -m simulink_cli --json '{"action":"schema"}'
  - Execute all tests via CLI commands
  - Generate test cases dynamically from schema output

ALLOWED:
  - Read skills/simulink_automation/SKILL.md (shipped skill — Phase 5 doc-accuracy target)
  - Read skills/simulink_automation/reference.md (shipped reference doc)
  - Read docs/reports/ (compare with previous reports)
  - Read git log (determine incremental scope)

Reading the shipped skill files is a deliberate, scoped exception — the purpose is to verify that documentation actually guides users to correct results.

Test Mode Selection

When invoked, determine the test mode:

Does docs/reports/LIVE-TEST-REPORT.md exist?
├─ No → FULL test (all phases, all actions)
└─ Yes → Read report, extract test_commit and schema snapshot
         Did user request "full"?
         ├─ Yes → FULL test
         └─ No → INCREMENTAL (see Incremental Logic below)

User specifies "retest FAIL-001" or "retest Phase 3"?
└─ TARGETED test (only specified items)

Incremental Logic

1. Read test_commit from existing report
2. Run: git log {test_commit}..HEAD --oneline
3. Run: python -m simulink_cli --json '{"action":"schema"}'
4. Compare schema output with Schema Snapshot section in report

Mandatory re-tests:
  - All items with status FAIL
  - All items with status BLOCKED (re-check environment)

AI-judged additions:
  - Commit messages reference specific actions → re-test those action's items
  - skills/simulink_automation/SKILL.md changed → re-test Phase 5
  - Schema output differs from snapshot → re-test Phase 1 + Phase 4
  - Optionally spot-check some PASS items for regression

If changes are large-scale (refactor, new actions added to schema):
  - Suggest full test to user, ask for confirmation

Execution Flow Overview

Read .claude/skills/live-testing/phases.md for detailed phase definitions before executing.

PhaseNameScope
0Environment CheckMATLAB reachability (always runs)
1Meta Layerschema, session, list_opened
2Read-Only Layerscan, find, connections, inspect, highlight
3Lifecycle & Writemodel_open, model_new, set_param, model_save
4Error HandlingInvalid inputs, error codes, rejection behavior
5DocumentationSKILL.md accuracy, schema-behavior consistency
6ReportCompile results → docs/reports/LIVE-TEST-REPORT.md

Action lists are illustrative for v2.2.0. When schema reveals unlisted actions, assign by characteristic: meta → Phase 1, read-only → Phase 2, lifecycle/write → Phase 3.

Incremental mode: skip unaffected phases, but Phase 0 always runs.

For report format and update rules, read .claude/skills/live-testing/report-template.md during Phase 6.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.