tailwind

Run independent adversarial participants through host-native sub-agents by default or separate model CLIs by choice — brainstorm, refine, plan, execute, architect, review, and synthesize

Generate, refresh, recut, and review gold-standard PR descriptions; detect AI slop, brittle counts, and verbose commit messages on branch commits and resolve via fixup commits and autosquash with quality-gate checks

Merge PRs via gh with merge commits matching the repo's own git history — readiness checks and CI watch before every merge, plus multi-PR runs with stack detection, rebase, and conflict resolution between merges

Cut and bump releases with safe defaults — no push, no tag, no tag push without an explicit flag — and roll new releases out to every downstream consumer repo with CI verification

Take tickets to branches in isolated git worktrees — resolve Linear/GitHub issues strictly read-only, name branches by the team's own conventions, land gated commits, and fan out multiple tickets in parallel

Measure and report the business value of AI skills and agentic workflows — provenance-tagged data collection and audience-tiered reports in engineer-hours and cycle time, never currency

Move Terraform and OpenTofu versions across any repository layout — discover every root module instead of assuming one, move every declaration of a provider together because constraints combine across modules, and refresh each lock file without narrowing its platform coverage

Prove a path fast in a no-commit spike — a single probe or a bakeoff of competing strategies in git worktrees — exit through the project's quality gates into stashes, and hand back a commit-by-commit implementation plan

Update GitHub Actions pins across one repo or a whole fleet — verify every target tag exists before writing it, research each upgrade against real release notes, land one commit per action on trunk, then close dependabot's PRs by citing the commit that superseded them

Make verification requests return the re-derived answer instead of a diff against the agent's prior turn — an ambient skill for 'double check' / 'are you sure' moments plus an /align command to repair a chat where the diff already happened