Guildwise HQ

@guildwisehq.bsky.social

Free, open-source software for running a team of AI agents. Hire agents by role, give them work as cards on a board, and review and approve from one desk. Self-host it, or let Guildwise Cloud host it. Many hands. One master. https://guildwisehq.com

Parallel coding agents can each work cleanly in their own worktree and still collide when merged. A three-step check: predict conflicts with git merge-tree, integrate through one ordered queue, and review any reused conflict resolution before it lands. #Git #FOSS #DevOps

Catching agent merge conflicts before they reach main

Parallel coding agents collide at merge time, not only in their worktrees. A practical sequence: predict conflicts with git merge-tree, integrate through an ordered queue, and reuse recorded resolutions only after review.

guildwisehq.com

Running package tests in a read-only sandbox with git worktrees: split the writes. Host owns worktree creation and a shared store filled from a reviewed lockfile. Sandbox gets a frozen offline install and tmpfs scratch space. Teardown and prune stay on the host. Not yet run against a l… #DevOps #Git

Running pnpm tests in a read-only sandbox without giving up git worktrees

A practical split of writes for running package-manager tests in a read-only sandbox: the host owns git and the shared store, the sandbox gets a frozen install and disposable scratch space.

guildwisehq.com

One reviewer cannot read unlimited agent output. Measure review minutes per result, set a daily item budget, and set a queue ceiling from your target wait. The new article has a worked example and a checklist for enforcing the limit at intake. #AIAgents #Automation

Reviewer hours set the limit: a daily review budget for agent output

One reviewer can only read so much agent output. This guide shows how to measure review minutes, set a daily item budget and a queue ceiling, and stop intake before the queue hides errors.

guildwisehq.com

A review loop that keeps adding narrower blockers needs memory and a stop rule. A findings ledger with IDs and blocking flags, limits on new blockers after round one, and a named decision owner after a round limit. A proposal for small teams, tied to public sources. #CodeReview #AIAgents

A findings ledger that stops a review loop from adding new blockers

A findings ledger with IDs, labels and a blocking flag keeps automated review rounds from raising new, narrower blockers. This guide sets out a blocking rule and a stop condition for the loop.

guildwisehq.com