Herdr team coordination policy for AI coding agents
93%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
The optimizer is a delivery specialty using the existing developer
responsibility. It adds no runtime role, scheduler, ledger or review gate.
When optimization is authorized, keep it as a standing delivery objective in
the current task/goal. Give an eligible developer a bounded area, improvement
goal, worktree, allowance, behavior constraints and normal report contract.
Include Skill(skill: "herdr-optimizer") in that worker's brief. Discovery and
implementation can share one assignment; do not require a preliminary audit.
Dispatch independent optimization slices alongside feature delivery. Use separate tasks/worktrees for separate writers. Avoid areas under active modification; retain dependencies and contributor exclusions. Select the minimum adequate configured worker through normal planning; discovery is not automatically a mechanical round. Reserve no permanent model, pane or subscription.
Handle each delivered change through ordinary independent verification and release. Continue the next worthwhile slice under the standing authorization without requiring another user request. No fleet-wide optimization pass, finding queue, line-deletion quota or per-change administrative review is required.
If the worker finds no worthwhile slice, let it finish and use capacity for other authorized work. Revisit optimization when relevant code, evidence or scope changes. Never keep an agent burning tokens to re-read unchanged code.
.tessl-plugin
rules
skills
herdr-foreman
classify
foreman
references
specialists
templates
herdr-observer
herdr-optimizer
herdr-standup