Be Review

Review gauntlet (serial)

Run the reviewers one after another on the live branch, each the sole editor while it runs. When a step starts the previous one has already committed, so every reviewer reads a clean, settled tree and applies its own fixes directly. Why serial, and the incidents behind the rules below: RATIONALE.md — read it when editing this skill, not when running it.

1. architecture-first-principles — FIRST, for a diff touching framework packages (@kolu/surface*), adding/reshaping module structure, or whose correctness rests on a concurrency / ordering claim (a race declared closed, one event asserted to land "before" another, timing across independent async cascades, a flag flipped by one callback that another path reads back as truth): run the named checks per their SKILL (Workflow fan-out, adversarial verify, scope = diff + one hop down its imports). Skip ONLY for pure-docs or trivially-local diffs — and a diff leaning on a happens-before is not trivially-local (RATIONALE.md). Say so either way. Its confirmed findings are dispositioned like any stage's — fix now or record where, never "acceptable for scope".
2. /lens-debate — lowy and hickey review boundaries/simplicity as two independent parallel subagents, then one reconcile-and-apply pass commits the fixes (grouped by area, minors batched). Pass the change rationale so the lenses don't flag deliberate decisions.
3. /agent-debate — the explicitly selected Claude, Codex, or Grok peer debates the current author agent to consensus. Its author rounds edit and auto-commit fix(…) on the branch.
4. /simplify — the self-applying reuse / simplification / efficiency pass over the changed code. Now that nothing runs concurrently, it runs as itself (it could not against the old read-only snapshot).
5. code-police — its rule-checklist and fact-check passes, applying their fixes. Run with --no-elegance so its elegance pass is skipped: that pass re-invokes /simplify, which step 4 already ran over this same tree.

Each step runs to completion before the next begins.

PR comments come after the push, never before. Each step commits locally but be-review pushes only once, after all selected steps finish, then posts. The debate skills run with their self-commenting suppressed (--no-comment) and each leaves its body on disk — the lens skill writes .lens-debate/comment.md; the agent-debate body is a compact commit table you assemble in step 2. No PR comment can reference a local-only commit.

Preflight

• Non-empty diff. git diff --stat <base> (default: the repo default via git symbolic-ref --short refs/remotes/origin/HEAD). If empty, stop.
• Commit first. Reviewers review committed code — commit/stash any outstanding work before starting (in /be this is automatic: §2/§3 commit and push before §4).
• Verify <base> is a remote-tracking ref — mechanically, not by eyeballing the diff stat. A caller's --base master names the local master, which on a long-lived worktree is routinely tens of commits behind origin, so the merge-base lands in the base branch's own past and the "diff" swells with history nobody wrote: one run resolved a base 40 commits stale and handed the reviewers 377 files instead of 9. After git fetch origin, run git -C "$repoPath" rev-parse --verify --quiet "refs/remotes/$base" — if it fails, retry once as origin/$base and use that; if that fails too, stop and say so rather than resolving a local ref. Never accept a bare branch name as given.
• Resolve the scope once. git fetch origin, then MB=$(git merge-base <base> HEAD) and START=$(git rev-parse HEAD). Pass MB as the base to every step (their own merge-base resolution is idempotent on a SHA) so each reviews the change against the identical fork point. Note that each step sees the commits the previous step added as part of the diff — that is intended: a later reviewer reviews the earlier reviewer's fixes too. Run every git here with git -C "$repoPath" (below) so a cross-repo run resolves the target repo's base, not the cwd's.
• Pin repoPath — the repo under review may NOT be the cwd. A /be run can carry the work in a companion repo (e.g. the drishti PR a @kolu/surface change requires per /be §5) while the session is rooted in a kolu worktree. Set repoPath to that target repo's absolute path (default: the cwd worktree root) and thread it into every step — every subagent prompt must carry it, with absolute paths and git -C "$repoPath", never a bare relative path (RATIONALE.md has the incident). If a cross-repo step returns clean against a non-empty target diff, suspect repoPath didn't take effect before trusting it.
• Debate peer (unless --tracks excludes debate): require --agent <claude|codex|grok> and run /agent-debate's matching auth preflight. If it fails, name the matching login command or exact non-auth error and stop the gauntlet; never skip a mandatory track or select a fallback peer.

Run the steps in order

--tracks checks,lens,debate,simplify,police selects which steps run (default all), in the listed order. Run each to completion, then move to the next. Preflight already ran git fetch origin and resolved the base, so pass MB straight into each step and skip the per-skill step-1 fetch / base resolution — don't redo it once per step.

Both review steps run inline, in your own turn. /lens-debate fans two lens subagents out in one message and merges their findings in a single reconcile-and-apply pass — no background Workflow to await. /agent-debate drives a live selected-peer split terminal that pings you each round — follow its event-driven loop to consensus directly. Act only on real signals (a subagent returning, a peer ping), never a timer: don't schedule ScheduleWakeup polls and don't babysit (a prior run wired 4-min wakeups and a 5-min /loop to nudge a gauntlet that was simply mid-review — pure churn).

1. lens — follow /lens-debate (Skill tool). repoPath = the live worktree, base = MB, --no-comment (so it doesn't advertise its local-only commits before be-review pushes — defer the comment until after the push), and thread the rationale through. It commits the agreed fixes and leaves its rendered comment at $repoPath/.lens-debate/comment.md — post that file after the push (gh pr comment <pr> -F …), don't re-improvise a summary. The lens findings and outcome sit beside it in the same directory if you need to check what a lens actually said.

   /lens-debate returns a status of clean, applied, needs-human, or merge-base-error:

   • clean / applied — the lenses' findings settled and the fixes are committed.
   • needs-human — one or more findings sit in the skill's valve 2 (a judgment-shaped call, or a contradiction that hit the debate cap). /be §4 requires you to adjudicate each one yourself before moving on: surface it in the report with both lenses' positions, decide drop or apply, and apply the survivors before continuing. Fold your adjudication into the deferred lens comment. Never report a needs-human run as settled.
   • merge-base-error — the scope couldn't be trusted; report it and move on.

2. debate — require the caller's explicit --agent <claude|codex|grok>. Capture DEBATE_START=$(git -C "$repoPath" rev-parse HEAD) first so the later comment includes only this debate's commits. Invoke /agent-debate as:

   review --agent <selected> --repo "$repoPath" --base MB
   --no-comment --context <context> --rationale <rationale>
   

   Pass no reasoning-effort flag — the peer runs at its own CLI default, per /agent-debate.

   --no-comment defers the comment until after the final push. --repo keeps every git/scratch/gh/spawn operation rooted in a cross-repo target. --context carries the task intent to the author; --rationale tells the independent peer which decisions are deliberate.

   When an API-facing shared-surface change trips the drishti companion-repo gate, add this to --rationale: the gate is deferred to §ship because it can be checked only against final post-gauntlet Kolu HEAD, so it is not a blocking code finding. The up-front rationale prevents any selected peer from debating an impossible mid-review gate.

   /agent-debate runs inline, driving the selected peer's live split and committing each author round locally. On consensus, assemble the compact deferred comment from the debate's commit range:

   workDir="$repoPath/.agent-debate"
   rounds=$(find "$workDir" -maxdepth 1 -type f -name 'verdict-[0-9][0-9][0-9].json' | wc -l)
   commits=$(git -C "$repoPath" rev-list --count "$DEBATE_START"..HEAD)
   [ "$rounds" -ge 1 ] || { echo "agent debate: no verdicts — no completed debate" >&2; }
   authorName="<actual invoking harness name>"
   peerName="<selected peer display name>"
   {
     printf '## %s ⇄ %s debate\n\n' "$peerName" "$authorName"
     printf '✅ Consensus in %s round(s)\n\n' "$rounds"
     if [ "$commits" -gt 0 ]; then
       printf '| Fix commit | SHA | Description |\n|---|---|---|\n'
       git -C "$repoPath" log --reverse --format='%h%x09%s' "$DEBATE_START"..HEAD \
         | nl -w1 -s$'\t' | while IFS=$'\t' read -r n sha subj; do
             printf '| %s | %s | %s |\n' "$n" "$sha" "$subj"
           done
     else
       printf '_No fix commits: the peer approved after author disputes._\n'
     fi
     printf '\n**Legend** — findings %s raised:\n\n' "$peerName"
     jq -rs '[.[].findings[]] | unique_by(.id) | sort_by(.id|ltrimstr("F")|tonumber)[]
             | "- **\(.id)** — \(.issue|split(". ")[0])"' "$workDir"/verdict-*.json
   } > "$workDir/comment.md"
   

   Fill authorName from the actual running harness and normalize the selected peer to Claude, Codex, or Grok; never hardcode either side. If the debate ends in reviewer-error or merge-base-error, do not create or post a false-consensus body. Report the selected peer with the failure, then move on (after fixing and rerunning a bad base).

3. simplify — invoke /simplify (Skill tool), scoped to the change vs MB. It applies its fixes to the working tree. When it finishes, commit what it changed (refactor: simplify <area>, staging only the files it touched). If it changed nothing, note that and move on.

4. police — invoke /code-police (Skill tool), passing --no-elegance whenever the simplify track (step 3) ran this gauntlet. That flag skips Pass 3 (elegance), which would otherwise re-invoke /simplify over the tree step 3 already simplified — a full skill invocation to re-derive a near-guaranteed no-op. Pass 1 (rules) and Pass 2 (fact-check) still run. Only omit the flag when --tracks excluded simplify — then no standalone simplify ran, and the elegance pass is the run's one simplify, not redundant. Its embedded pass prompts diff against origin/HEAD...HEAD by default, which is wrong whenever --base isn't the repo default: before invoking, tell the police passes to scope to MB — pass the merge-base explicitly so every pass runs git diff <MB>...HEAD, not the default ref. Apply the fixes it surfaces, committing each fix(police): <title> with the finding in the message (stage only the files changed).

Push, then comment

First settle whether there is anything to push: git log --oneline $START..HEAD ($START was captured in Preflight). Then:

• Run just fmt before any push, whenever new commits exist. No reviewer guarantees formatting and just check never runs the formatter, so an unformatted tree sails through a green check and reds ci::fmt later (RATIONALE.md). just fmt runs both biome-format and nixpkgs-fmt — the only gate matching ci::fmt. If it changed anything, commit the reformat (style: just fmt, staging only what it touched) so it rides this push.
• New commits exist and a PR exists for this branch (gh pr view --json number -q .number) → git push. Only after the push succeeds do you post the deferred comments — the lens and agent-debate bodies from steps 1–2 are now safe to publish because the SHAs they name are on the remote.
• No new commits (every step was clean or applied nothing) but a PR exists → there is nothing to push, and HEAD is already remote-visible, so post the deferred comments immediately. The local-only-SHA invariant is about never advertising an unpushed commit; with no new commit there is no such risk.
• No PR → there is nothing to push to and nothing to comment on. Skip both; the local commits (if any) and their findings live in chat and the local log for the human.
• A required push fails → do not post the comments (the SHAs are still local-only); report the push failure instead.

Never merge — pushing updates the open PR; the human reviews the commits and merges when satisfied.

When you do post, post one comment per track that produced a body — skip any track --tracks excluded, and skip a track that ran but yielded no postable comment (lens on merge-base-error, debate on persistent reviewer-error): the lens body and the agent-debate body verbatim (gh pr comment -F — the debate body is the $workDir/comment.md you assembled in step 2 as a compact commit table), and the police summary (the ## [👮 Code-police](https://agency.srid.ca/) comment described in Report).

Report

Summarize in chat — reporting only the selected tracks, and naming any track --tracks skipped so the absence is explicit, not silent:

• lens — status (applied + the fixes, or needs-human + which findings needed adjudication and how you adjudicated each, or merge-base-error); its PR comment landed (posted after the push) — except on merge-base-error, which has no comment body to post.
• debate — selected peer + consensus / reviewer-error; on consensus its PR comment landed (posted after the push, per "Push, then comment") — on reviewer-error there is no comment to post.
• simplify — whether it changed anything and what it committed.
• police — findings and how each was actioned; the ## [👮 Code-police](https://agency.srid.ca/) summary comment landed (posted after the push, alongside the lens and agent-debate comments).
• whether the fixes were pushed;
• git log --oneline <base>..HEAD + git diff --stat <base> so the combined result is visible.

ARGUMENTS: