fix: return action results from action_invoke

[Adkid-Zephyr]

Summary

• invoke actions through the OpenServerless CLI's result-returning ops invoke alias
• return the action's JSON response to the MCP client instead of an asynchronous activation ID
• align the MCP description, README, and specification with the actual command
• add regression coverage at the spawned ops process boundary

Root cause

action_invoke used ops action invoke. In the OpenServerless CLI, action is the generic wsk action wrapper, so this path invokes with blocking=false&result=false. Its successful output contains only an activation ID.

The CLI's dedicated invoke alias expands to wsk action invoke -r, which invokes with blocking=true&result=true and prints the action result.

Because this MCP exposes no follow-up activation-result tool, the old success response was a dead end for an agent: the action ran, but its result was unavailable.

Live verification

I exercised the checksum-verified Apache OpenServerless CLI release v0.9.1-2607122150.dev against a local mock OpenWhisk API:

• before: blocking=false&result=false; MCP output ok: invoked /_/v1/demo with id activation-123
• after: blocking=true&result=true; MCP output { "answer": 42 }

The request body remained { "key": "value" }, confirming parameter forwarding is unchanged.

Validation

• npm test — 14 tests passed
• npm run typecheck
• git diff --check
• independent structured review — no actionable findings

AI assistance

This contribution was developed with OpenAI Codex assistance. The submitter verified the upstream CLI contract, live behavior, patch, and tests. The commit includes a Generated-by: OpenAI Codex trailer.

[Adkid-Zephyr]

Hi maintainers, a gentle follow-up on this PR. It fixes the default action_invoke path so callers receive the invoked action result, aligning the implementation with the documented tool contract, and includes a process-boundary regression test. The license workflow is currently awaiting approval to run. I am happy to address any feedback or adjust the approach. Thanks!