Evaluate whether a harness change to the LLM Workbench is an improvement and run its feedback loop: claims to test, evaluation design and commands, the harness feedback loop, the automated feedback gate, automation run outcomes and manual harness feedback reports. A maintainer skill of this repository only; it never…
Cut, prove and publish an LLM Workbench release from this producer repository: freeze version labels, upgrade the reference Template, prove the composed round trip, portability and cross-provider resume, derive a room from recorded decisions and publish to the personal catalog. A maintainer skill of this repository…
Check how this producer repository lays out, installs, adopts, upgrades and reports on a room: the skills lane, support root, managed runtime tools, lifecycle classification, adoption, control fidelity, explicit upgrade and self-drift checks, plus the carrier line-landing, GitHub binding and socket contract checks. A…
Operate the Workbench runtime tools a room has installed - read a finding and what it blocks, validate the Wiki, repair installed state, allocate a visible identifier, keep the connection identity, check a configured host and add a room-local skill - through the procedure the room's Runbook operations index points to.
Implement a sliced Workbench Spec and prepare its assembly-to-integration PR for review.
★not rated 2 2d agoA20 tokens
originalMIT
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: