メニュー
Smoke test a Covia venue by running basic operations — echo, LLM, grid. Use to verify a venue is healthy and adapters are working.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Create, configure, and manage Covia agents. Handles config gotchas, system prompts, LLM backend setup, and lifecycle operations. Use when working with agents on a venue.
Demo cross-venue federated operations — invoke operations on remote venues, show distributed audit trails. Use for demonstrating Covia's federation capabilities.
Issue and manage UCAN capability tokens — delegate access to workspace, agents, secrets, and operations. Use for demonstrating or configuring Covia's authorisation model.
Set up and run the AP Invoice Audit Trail demo. Creates three agents (Alice, Bob, Carol) on a Covia venue, seeds reference data, and runs an invoice through the pipeline. Use when the user wants to demo or test the AP workflow.
Manage secrets on a Covia venue — store API keys and credentials. Use when setting up LLM providers or other integrations that need secrets.
Store, retrieve, list, and manage content-addressed assets on a Covia venue. Assets are immutable resources — operations, artifacts, agent definitions, orchestrations — identified by the SHA3-256 hash of their metadata. Use when working with any venue asset.
| name | grid-test |
| description | Smoke test a Covia venue by running basic operations — echo, LLM, grid. Use to verify a venue is healthy and adapters are working. |
| argument-hint | ["quick|full"] |
Prerequisite: The venue must be running and connected as an MCP server (http://localhost:8080/mcp). If MCP tools are not available, tell the user to run /venue-setup local first.
Run a quick health check on the connected Covia venue.
quick (default)Run these in parallel and report results:
Echo test — verify basic operation execution:
grid_run operation=test:echo input={"ping": "pong"}
Expected: returns {"ping": "pong"} unchanged.
List adapters — verify all adapters are registered:
covia_list path=v/info/adapters
Expected: should show adapters (test, agent, llmagent, grid, langchain, convex, http, jvm, covia, secret, ucan, mcp, orchestrator).
List operations catalog — verify operations are installed:
covia_list path=v/ops
Expected: 40+ operations across all adapters under their per-adapter sub-namespaces.
Report a summary table: adapter count, function count, echo test pass/fail.
fullRun all quick tests plus the tests below. Requires OPENAI_API_KEY — if not set, tell the user to run /secret set OPENAI_API_KEY first.
LLM test — verify OpenAI connectivity:
grid_run operation=langchain:openai
input={ "messages": [{"role": "user", "content": "Say hello in exactly 3 words"}] }
Expected: assistant message response.
Agent lifecycle — create, request, query, delete:
agent_create agentId=_smoke_test config={"operation": "v/test/ops/taskcomplete"}
agent_request agentId=_smoke_test input={"test": true} wait=true
agent_info agentId=_smoke_test
agent_delete agentId=_smoke_test remove=true
Expected: agent runs, task completes, timeline has one entry.
Workspace CRUD — write, read, delete:
covia_write path=w/_smoke_test value={"ts": <now>}
covia_read path=w/_smoke_test
covia_delete path=w/_smoke_test
Expected: write succeeds, read returns value, delete succeeds.
Report a full summary table with pass/fail for each test and any errors encountered.