kmp-ledger

Determine the latest Kotlin + AGP versions compatible with BOTH the current Android Studio and the current IntelliJ IDEA (the lagging IDE is the cap), compare against this project's pins, and present a dated inline report. Researches live; report-only — proposes a bump but never edits without confirmation. Do not invoke automatically.

Cut a new release — analyse source changes, recommend the next version, confirm it, then bump gradle.properties, write CHANGELOG.md, and commit. Optionally invoke with a version number to pre-fill the recommendation. Do not invoke automatically.

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.

Orchestrate a currency audit of the whole project against the latest official upstream best practices, using the nine bp-* specialist subagents dispatched in parallel waves, then synthesize one prioritized, source-cited review document written to docs/currency-review-YYYY-MM-DD.md (findings + a phased fix plan). Writes only that doc; makes no other code changes (each subagent may persist notes to its own project memory). Do not invoke automatically.

Orchestrate a full house-rules review of the entire project from the eleven rv-* review specialists, dispatched in three parallel waves, then synthesize one prioritized review document written to docs/house-review-YYYY-MM-DD.md (findings + a phased fix plan). Writes only that doc; makes no other code changes (each subagent may persist notes to its own project memory). Do not invoke automatically.

Audit this project's `.claude/` subagents, skills, and hooks against the LATEST official Claude Code docs, then present a dated best-practices report inline. Researches current guidance live (does not rely on memory). Read-then-offer — proposes fixes but does not apply them without confirmation. Do not invoke automatically.

Analyse source changes since the last release and produce a plan describing what in README.md needs updating. Does not edit any files. Do not invoke automatically.

Review all staged changes and create a conventional commit message derived from the actual diff. Use after running git add. Do not invoke automatically.