The seven steps of one judge round, run in order from skills/herdr-foreman/SKILL.md Step 13. Modes, triggers and both report contracts are in skills/herdr-foreman/references/round-flow.md "The Judge".

Each command resolves CP to the local or home plugin, or to a coding-policy clone, the same way SKILL.md does. Repeat its resolver in every call.

1 — Compose the Judge Brief

Optional. Modes, triggers and both report contracts are in skills/herdr-foreman/references/round-flow.md "The Judge" (a bot disagreement inside SKILL.md Step 14 returns here first). No trigger — proceed to SKILL.md Step 14.

For a dispute, compose from templates/brief-judge.md through SKILL.md Step 7: the dispute, both positions with report paths, the governing rule, the tree. Fill POSITION_A_EVIDENCE and POSITION_B_EVIDENCE with the citations each report makes (file and line, or command output, at a revision), copied, never supplied by the foreman. A position that cites nothing is not ready for the judge: dispatch an investigator under skills/herdr-foreman/references/specialists.md to establish the disputed facts with citations first, fill that position's evidence value with supplied by the investigator report, and fill INVESTIGATION_REPORT with that report. Fill INVESTIGATION_REPORT with the investigator's report after an insufficient ruling as well, otherwise "none".

For an exhausted allowance, compose from templates/brief-judge-diagnosis.md through SKILL.md Step 7 under the role key judge-diagnosis, which writes brief-judge-diagnosis.md: the assessed investigator report, the task, rounds spent, remaining blocking work, the per-round history, the tree, and any prior remedy with what it changed. The pinned seat is still judge, so plan and dispatch that role and pass this file as its brief: --brief judge=<outdir>/brief-judge-diagnosis.md.

For a weighing, compose from skills/herdr-foreman/templates/brief-judge-weighing.md under the role key judge-weighing, and pass --brief judge=<outdir>/brief-judge-weighing.md. One weighing carries every nomination at one gate. Nominate from two sources only: a worker report's MARGINAL: line, and the churn signal below. Leave out a finding a ruling already covers. For internal findings --from is the previous record-report receipt's head_revision; for policy findings it is the policy reviewer's previous review commit_id. The initial implementation round has no --from and no churn signal.

CP=.tessl/plugins/jbaruch/coding-policy; [ -d "$CP" ] || CP="$HOME/$CP"; [ -d "$CP" ] || case "$(git config --get remote.origin.url)" in git@github.com:jbaruch/coding-policy|git@github.com:jbaruch/coding-policy.git|https://github.com/jbaruch/coding-policy|https://github.com/jbaruch/coding-policy.git|ssh://git@github.com/jbaruch/coding-policy|ssh://git@github.com/jbaruch/coding-policy.git) CP=. ;; *) echo "coding-policy plugin not found: run tessl install jbaruch/coding-policy" >&2; exit 1 ;; esac
bash "$CP/skills/herdr-foreman/foreman.sh" finding-churn \
  --repo <shared-checkout> --from <previous-head> --to <head> \
  --finding <path>:<line> [--finding <path>:<line>]...

Emits {from, to, findings: [{path, line, added_by_last_fix, path_changed}]}. A finding with added_by_last_fix true is nominated. Exit 1 names a bad reference or finding on stderr; fix it and rerun. Fill NOMINATIONS with each nominated finding verbatim, its source, path:line and rule copied exactly, and its evidence; policy findings come from the list-mode JSON in the release worker's report. HEAD is that head's full sha. Fill INVESTIGATION_REPORT after an insufficient weighing, otherwise "none".

Skip SKILL.md Step 8 for the read-only judge. Proceed immediately to step 2.

2 — Re-measure the Shared Window

Re-run SKILL.md Step 4's measure under its outcome contract. Resolve any unreadable judge window before planning. Pass the fresh snapshot to step 3; never reuse the earlier reading as affordability proof. Proceed immediately to step 3.

3 — Plan the Judge Seat

Plan the pinned judge against step 2's fresh snapshot:

CP=.tessl/plugins/jbaruch/coding-policy; [ -d "$CP" ] || CP="$HOME/$CP"; [ -d "$CP" ] || case "$(git config --get remote.origin.url)" in git@github.com:jbaruch/coding-policy|git@github.com:jbaruch/coding-policy.git|https://github.com/jbaruch/coding-policy|https://github.com/jbaruch/coding-policy.git|ssh://git@github.com/jbaruch/coding-policy|ssh://git@github.com/jbaruch/coding-policy.git) CP=. ;; *) echo "coding-policy plugin not found: run tessl install jbaruch/coding-policy" >&2; exit 1 ;; esac
bash "$CP/skills/herdr-foreman/foreman.sh" plan \
  --roles judge --judge-mode <adjudication|diagnosis> \
  --snapshot <step-2-measure-output> --task <task-id>

adjudication for a dispute or a weighing, diagnosis for an exhausted allowance — the same choice step 1 made when it composed the brief. Use the recorded mode in steps 4 and 5. An undeclared mode is refused.

Exit 0 names the judge worker; proceed immediately to step 4. On non-zero, report the diagnostic and finish here. Never substitute a judge, lower its tier, or hand-write an assignment to bypass the refusal.

4 — Start the Judge Worker on Its Pinned Tier

For an existing judge worker, proceed to step 5 with a clearing dispatch. For an empty shell pane, complete retrospective checks for the start and run:

CP=.tessl/plugins/jbaruch/coding-policy; [ -d "$CP" ] || CP="$HOME/$CP"; [ -d "$CP" ] || case "$(git config --get remote.origin.url)" in git@github.com:jbaruch/coding-policy|git@github.com:jbaruch/coding-policy.git|https://github.com/jbaruch/coding-policy|https://github.com/jbaruch/coding-policy.git|ssh://git@github.com/jbaruch/coding-policy|ssh://git@github.com/jbaruch/coding-policy.git) CP=. ;; *) echo "coding-policy plugin not found: run tessl install jbaruch/coding-policy" >&2; exit 1 ;; esac
bash "$CP/skills/herdr-foreman/start-judge-worker.sh" \
  <step-3-plan-file> <pane> [claude|codex|grok] --task <task-id> [--state <state-file>]

Starts the pinned judge and verifies launch argv, on the mode step 3 recorded in the plan. The header owns the contract.

Exit 0 — proceed immediately to step 5 with --no-clear.
Any non-zero — report the diagnostic and finish here without briefing the worker or overriding its tier.

5 — Dispatch the Judge

Use step 3's plan under SKILL.md Step 10's dispatch contract:

CP=.tessl/plugins/jbaruch/coding-policy; [ -d "$CP" ] || CP="$HOME/$CP"; [ -d "$CP" ] || case "$(git config --get remote.origin.url)" in git@github.com:jbaruch/coding-policy|git@github.com:jbaruch/coding-policy.git|https://github.com/jbaruch/coding-policy|https://github.com/jbaruch/coding-policy.git|ssh://git@github.com/jbaruch/coding-policy|ssh://git@github.com/jbaruch/coding-policy.git) CP=. ;; *) echo "coding-policy plugin not found: run tessl install jbaruch/coding-policy" >&2; exit 1 ;; esac
bash "$CP/skills/herdr-foreman/foreman.sh" apply \
  --assignments <plan-file> \
  --brief judge=<round>-judge.md --report judge=<absolute-report-path> \
  --common <path-to-COMMON.md> --judge-mode <adjudication|diagnosis> \
  --task <task-id> [--no-clear]

Pass the same --judge-mode step 3 planned. SKILL.md Step 10's outcomes govern. Use --no-clear only for the worker just started in step 4; an existing judge receives the default cleared relaunch with retrospective coverage. Apply verifies the live tier before input. Proceed immediately to step 6.

6 — Wait for the Ruling

Run SKILL.md Step 11's fleet observation loop, including the judge named by step 3. Proceed immediately to step 7 once its report lands; keep other enrollments under observation.

7 — Act on the Ruling or Remedy

For a weighing, verify the delivered report before recording or applying any of its outcomes:

CP=.tessl/plugins/jbaruch/coding-policy; [ -d "$CP" ] || CP="$HOME/$CP"; [ -d "$CP" ] || case "$(git config --get remote.origin.url)" in git@github.com:jbaruch/coding-policy|git@github.com:jbaruch/coding-policy.git|https://github.com/jbaruch/coding-policy|https://github.com/jbaruch/coding-policy.git|ssh://git@github.com/jbaruch/coding-policy|ssh://git@github.com/jbaruch/coding-policy.git) CP=. ;; *) echo "coding-policy plugin not found: run tessl install jbaruch/coding-policy" >&2; exit 1 ;; esac
bash "$CP/skills/herdr-foreman/foreman.sh" verify-ruling \
  --task <task-id> --ruling <absolute-report-path>

Exit 0 prints {task, dispatch, judge, report, sha256}: the report is the one the pinned judge's weighing dispatch was enrolled to write, from the weighing brief. Any non-zero exit applies nothing: record no defer or decline, keep every nominated finding blocking, and report the diagnostic. The binding rules are the command's; see cmd_verify_ruling in skills/herdr-foreman/foreman/cli.py.

Coverage, from then on: a defer or decline finding stays covered while foreman finding-churn --from <ruling HEAD> --to <tip> (step 1's command) reports its path_changed false. List each covered finding as covered in scoped re-check and release briefs, and one whose path changed as no longer covered; it blocks again.

Apply the Ruling Outcomes contract in skills/herdr-foreman/references/round-flow.md. Investigation rulings return to SKILL.md Step 12's knowledge gate. Implementation rulings route unchanged-branch rulings to verified release or renewed verification, branch-changing rulings to the counted correction path, an insufficient ruling to an investigator and then back to step 1, and a blocked ruling to its saved operator question. Only the operator overrides a completed ruling; insufficient binds nothing and creates no checkpoint. A judge round is a round: log it in SKILL.md Step 16 and reset in Step 17, recording the step that outcome names as the stow's continuation step. The next context takes SKILL.md Step 17's Resume Route.