Convert planning state into autonomous build, test, and verification execution.
npx skills add https://github.com/lirrensi/agent-sommelier --skill engage-lirrensi
Please help me install this Agent Skill. Skill: engage Source: https://github.com/lirrensi/agent-sommelier/tree/main/skills/engage Command: npx skills add https://github.com/lirrensi/agent-sommelier --skill engage-lirrensi
It prevents an AI coding agent from stopping early or returning incomplete, unverified work by enforcing an autonomous execute-and-verify mode when the user is ready to act.
Example use case: You have an implementation plan for a software change and want the agent to complete the work end-to-end—without further clarification—while producing verifiable results you can inspect.
Tell your agent: Engage and execute the plan end-to-end with build and test verification, then deliver the verified outcome.
High-intent search queries and answers about installing and using this skill.
To stop an AI coding agent from stopping early, you must enforce an autonomous execute-and-verify mode. This requires the agent to produce concrete verification evidence, like passing build and test checks, before declaring any software delivery task complete.
Proof of work in autonomous software execution is mandatory evidence that an agent completed its planned tasks. It requires running concrete quality gates, such as build and runtime checks, and reporting verifiable success artifacts before finishing the workflow.
To execute an implementation plan end-to-end, you provide an explicit engage signal to freeze clarification requests. This switches the agent into an uninterrupted build, test, and verify mode that follows the plan-defined scope and delivers the verified outcome.
An autonomous agent can stop execution if it hits a catastrophic unrecoverable blocker. Safety guardrails allow stops for these severe issues while prohibiting early exits, hand-waving, and unnecessary questions during the software delivery workflow.
Autonomous plan execution does not work without a predefined software delivery plan. The process strictly requires an existing planning state to convert into an execution workflow, ensuring the agent follows a plan-defined scope to produce verifiable results.
FernanSuozaDefine autonomous execution boundaries and audit evidence for unattended AI runs.
JaimeJunrRun verification commands and report evidence before claiming task completion.
yutnaRun type checks, linting, tests, and verification handoff in canonical order.
twjarviszyra-webCoordinate implementation plans by dispatching independent subagents for each task.
