CtrlK
BlogDocsLog inGet started
Tessl Logo
PluginPublic

jbaruch/herdr-policy

Herdr team coordination policy for AI coding agents

74

Quality

93%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Overview
Quality
Evals
Security
Files

release-integration.mdskills/herdr-foreman/references/

Team Release Integration

This contract governs verified Herdr team rounds. General review, release and publication requirements come from the selected coding-policy release skill.

Team Weighing

With HERDR_ENV set, the release worker asks nobody and nominates nothing:

  1. Report the blocking findings to the foreman under ## BLOCKED, with the JSON dismiss-ruled-review.sh prints without --ruling.
  2. The foreman nominates from them; the pinned judge weighs the nominations, and its weighing report is the ruling file.
  3. The next release brief names that report, the follow-up issue and the task. Run the team-round invocation from Step 6, with --task; the script binds the report to the judge's weighing through foreman verify-ruling.

The question and the standalone ruling file below do not apply in a team round.

The generic dismiss-ruled-review.sh consumes the report through the optional team-ruling interface. Its owner verification must succeed before the parsed bytes can authorize dismissal. A missing Herdr package or unknown verification result never degrades to an operator ruling.

Review Exceptions

  • A weighing asks whether fixing a finding is worth its cost, never whether the finding is real
  • The ruling file follows the format the skills/release/dismiss-ruled-review.sh header names
  • In a Herdr team round the pinned judge weighs under skills/herdr-foreman/references/team-operation.md Judge Seat
  • The question never blocks the round
  • The agent keeps fixing until the answer arrives
  • No answer means fix
  • Preconditions (all required):
    1. The ruling rules the finding defer or decline
    2. In standalone mode, the ruling file carries AUTHORITY: operator and quotes the operator's answer verbatim
    3. In a team round, the ruling file carries AUTHORITY: judge and foreman verify-ruling binds it to the report supervision enrolled for the pinned judge's weighing
    4. The ruling names the finding at its HEAD:, and the finding's file is unchanged from that commit to the head
    5. The finding is under no floor
    6. The finding is entered in the task's follow-up issue citing the ruling, a decline labelled won't-fix
    7. Every other release gate holds
  • Floors:
    • a failing test, lint, diagnostic or required check
    • a security or data-loss defect reachable by normal inputs
    • a no-secrets or ci-safety finding
    • a carve-out's unmet precondition
    • an unmet acceptance criterion the operator stated
  • The judge rules fix on every floor
  • skills/release/dismiss-ruled-review.sh refuses the rule-id floors and a failing check
  • A ruling never skips, disables or removes a test
  • A gating policy review the ruling covers is dismissed per rules/ci-safety.md Judge-Ruled-Review Dismissal Carve-Out

Source Changes

The release worker reports source defects under ## BLOCKED; the foreman routes authorized corrections to their writer. Release workers never edit source to satisfy a review. Advisory comments need no follow-up issue or reference and never enter the merge predicate.

skills

herdr-foreman

compose-briefs.sh

config.example.json

foreman-tier-check.py

foreman.sh

label-workspaces.sh

native-hook.sh

policy-paths.py

provision-worktree.sh

prune-report-caches.py

resolve-gates.sh

resolve-policy-paths.sh

review-package.sh

roster.sh

round-preflight.sh

SKILL.md

start-judge-worker.sh

state-schema.md

sweep-worktrees.sh

verify-authority.sh

wait-report.sh

README.md

tile.json