SRI: The runtime component of BuLLM
Skills Runtime Implementation is designed to take an agreed workflow from request to reviewed result. Its role is runtime execution: using the right skill, working within its permissions and involving people at the points where a decision is required.
Start with a defined skill
A reusable skill needs a purpose, required inputs, permitted tools, expected output and acceptance checks. It also needs to say what should happen when information is missing or a step cannot be completed. SRI is intended to use that defined scope rather than treat every request as an open-ended task.
Make the execution sequence visible
The intended sequence is to select the relevant skill, check its requirements, prepare the steps, use the permitted tools and return the output. A step that needs approval should reach the named reviewer before the action proceeds. The deployment must implement and test those boundaries.
Give each role clear responsibilities
The workflow should identify who may run a skill, who reviews an exception and who manages its configuration. Writing down those roles is only the start: access and approval behaviour must be checked in the actual environment before the team relies on them.
Plan for exceptions and handoff
Define the evidence needed to review a run, including the selected workflow version, important decisions and unresolved errors. Recovery behaviour depends on the action and tools involved; not every action can be retried or undone safely. The handoff should state what completed, what did not and who acts next.
Confirm the deployment scope
This page describes the intended runtime role of SRI. Production readiness, a particular data-hosting arrangement and the available recovery controls need to be confirmed for the proposed deployment.
Tell us which workflow you want to run and where permissions, approvals and failure handling matter.