| name | receiving-code-review |
| description | Consumes `codeReview.findings[]` from the previous CODE_REVIEW step and dispatches per-domain fix agents through a 7-step escalation ladder. Each fix agent receives the finding body, the source file, browzer deps + mentions, and the relevant skill from `finding.assignedSkill`. Findings successfully resolved end as `status: fixed`; findings that exhaust all fix attempts are recorded as `status: tech_debt`. Use after `code-review` and before `write-tests`. Triggers: receive code review, apply code review fixes, fix the findings, close the review, fix-findings, address review feedback, resolve code review. |
| argument-hint | <featureId> |
You are a fix-dispatch controller. Close every finding from CODE_REVIEW on a 7-step model-escalation ladder.
Read context
!`browzer get-step CODE_REVIEW --id $ARGUMENTS`
$ARGUMENTS is the feature id passed by the orchestrator (e.g. feat-20260507-preamble-staging-migration); it is also the directory name under docs/browzer/.
The blob lists every finding with severity, file, deps, mentions, and assignedSkill. Process highest severity first.
Iteration ladder (per finding)
The full step-by-step ladder lives in references/iteration-ladder.md. Summary:
- sonnet attempt
- sonnet retry with explicit failure context
- sonnet + research dispatch (WebSearch / Context7)
- opus attempt
- opus retry
- opus + research dispatch
- tech-debt log (only after the prior six exhaust) — record under
finding.tech_debt: true
Haiku is forbidden for fix dispatch. Zero-tech-debt is the default; reaching step 7 requires recorded justification.
Spawn each fix attempt with subagent_type: browzer:fixer. For ladder steps 1–3 use model: sonnet; for steps 4–6 use model: opus. Set effort: xhigh for high-severity findings; effort: high for medium and low.
Per-finding output
Each fix agent writes:
docs/browzer/<feat>/staging/RECEIVING_CODE_REVIEW.<finding-id>.json
Shape reference: see template.md (auto-generated from the workflow CUE schema). Do not paste schema-claiming JSON into this body.
Aggregator (final)
After all findings are processed, write the merged file:
docs/browzer/<feat>/staging/RECEIVING_CODE_REVIEW.json
Shape reference: see template.md (auto-generated from the workflow CUE schema). Do not paste schema-claiming JSON into this body. Any field not present there is dropped on save.
The autosave hook validates and persists RECEIVING_CODE_REVIEW.json.
Persistence
The autosave hook persists staging/RECEIVING_CODE_REVIEW.json automatically on write. Recommended flags when manually invoking save-step:
--quiet --await — RECEIVING_CODE_REVIEW is load-bearing: write-tests reads it back immediately after this phase completes.
On validation failure, re-run with --hint-fixes for worked examples of valid values.
Done when
- Every finding has a per-finding JSON file.
- The aggregated JSON has
summary.fixed + summary.techDebt == total findings.
- Tech-debt entries each carry an
iterations[] of length 6 documenting the exhausted ladder.
Return one line: receiving-code-review: <fixed> fixed, <techDebt> tech-debt; <totalIterations> iterations.
