| name | review-all |
| description | Run BOTH review lenses over the whole project — the eleven rv-* house-rules specialists and the nine bp-* currency specialists — in parallel waves, then synthesize ONE prioritized, cross-deduplicated review document written to docs/full-review-YYYY-MM-DD.md (findings + a phased fix plan). Heavyweight (twenty subagents); for release gates / periodic full audits, not per-diff. Writes only that doc; makes no other code changes. Do not invoke automatically. |
| disable-model-invocation | true |
| allowed-tools | Read, Grep, Glob, Bash, Agent, Write |
Run a combined full-project review across both lenses at once: the eleven rv-*
house-rules specialists (.claude/agents/review/) and the nine bp-*
upstream-currency specialists (.claude/agents/currency/), then merge everything into
one prioritized, source-cited, cross-deduplicated review document. The skill's sole
output is that document — docs/full-review-YYYY-MM-DD.md, holding the findings and a
phased plan to implement the fixes; it makes no other code changes.
This is the union of /review-house ("does the code obey this project's own
rules?") and /review-currency ("do the code and our rules still match the latest
official upstream guidance?"). Use it for a release gate or a periodic full audit —
not on every diff. It dispatches twenty subagents; reach for the single-lens
siblings or the bundled /code-review when you don't need the full sweep. Follow the
shared orchestration rules in .claude/REVIEW-CONVENTIONS.md.
What it runs
- House (
rv-*, no web): rv-arch, rv-kmp, rv-concurrency, rv-compose,
rv-data, rv-di, rv-testing, rv-security, rv-perf, rv-build, rv-ci.
- Currency (
bp-*, web): bp-kotlin, bp-kmp, bp-compose, bp-room,
bp-koin, bp-gradle, bp-ci, bp-android, bp-testing.
See .claude/agents/README.md for the per-domain pairing matrix.
Waves
Run the house waves first, then the currency waves (keep the lenses separate so
each synthesis stays tractable), each wave in a single message carrying the scope, and
synthesize only after all twenty return:
- House — Wave H1 structural:
rv-arch, rv-kmp, rv-di, rv-build; Wave H2
behavioral: rv-concurrency, rv-data, rv-compose; Wave H3 cross-cutting:
rv-testing, rv-security, rv-perf, rv-ci.
- Currency — Wave C1 language + UI:
bp-kotlin, bp-kmp, bp-compose; Wave C2
data + DI + build: bp-room, bp-koin, bp-gradle; Wave C3 platform + supply +
test: bp-ci, bp-android, bp-testing.
Findings-discipline rule (end every spawn's prompt with this)
Use each agent's family rule verbatim: the rv-* spawns get the house rule (report
only correctness / project-rule gaps; severity, file:line, concrete fix), the
bp-* spawns get the currency rule (report only upstream-currency gaps; severity,
file:line, fix, and source URL + version/date; defer internal correctness to the
rv-* pair). Neither invents findings when an area is sound.
Synthesize — one review document
Merge all twenty reports and write docs/full-review-YYYY-MM-DD.md (review date)
per the Synthesize — write the review document section of
.claude/REVIEW-CONVENTIONS.md. The doc has two parts: (1) findings in the three
tiers (Critical / Should-fix / Optional), and (2) a phased implementation plan for
the fixes (phases ordered by risk/value, each committable, with files + a ./gradlew
verify step and any apiDump/Kover gates).
Apply the Cross-skill de-duplication rule in .claude/REVIEW-CONVENTIONS.md: a
paired domain that surfaces the same file:line from both its rv-* and bp-* agent
becomes one entry — rv-* owns the correctness verdict, bp-* carries the source
citation. Include the pinned version vs. latest stable currency table from the
currency lens. Writing that one doc is the only file change; then offer next steps
(including implementing the fixes by phase) as described in
.claude/REVIEW-CONVENTIONS.md.