alonbaron

architect

Produces Alon's design-doc system BEFORE code: SOURCE_OF_TRUTH, ARCHITECTURE_ROADMAP, TODO_WORKFLOW, CLAUDE.md (+ modular docs/architecture). Model first: data + invariants → core enforcement → failure paths → API contracts → phased workstream. Use proactively when a new app, feature, or workstream is starting and code hasn't been written, or docs may have drifted (audit). Also on "architect", "design doc", "spec this out", "roadmap". Not for small fixes inside a current design.

ask-the-council

Convenes a panel of opinionated advisors (parallel Claude subagents, each with a distinct mandate and a forbidden move so they genuinely diverge), then a Chairman synthesis that COMMITS to one recommendation with explicit tradeoffs. Use proactively for any high-stakes design, architecture, or tradeoff decision where credible options genuinely compete. Also on "ask the council", "the council", "get a panel". Not for fact-finding or decisions with one obvious answer — answer those directly.

ponytail

Forces the laziest solution that actually works — YAGNI, reuse before new code, stdlib before custom, native platform before dependencies, one line before fifty. Levels: lite/full/ultra. Use proactively when a solution is growing beyond the minimum: new abstractions, dependencies, scaffolding "for later", or complaints about over-engineering or bloat. Also on "ponytail", "be lazy", "yagni", "simplest solution", "do less". Not at trust boundaries, not when the full version is explicitly asked, and not for tidying an already-written diff — that is built-in "simplify". Ponytail governs what gets built, before and while it is written; simplify cleans up after.

prompt-generator

Turns a vague ask into a rigorous, grounded, token-efficient prompt: role + objective, testable done criteria, anti-hallucination (verify or say "I don't know"), anti-tokenmaxing (lead with the answer, output budget), strict agent discipline. Use proactively when the user hands over a prompt or task spec destined for another agent or LLM. Also on "write a prompt", "improve this prompt", "prompt for an agent". Not for human-facing prose or work you'll execute yourself this session.

Code & Development Listed

review-swarm

Local, free, multi-specialist review of a diff: parallel Claude subagents (correctness, security/trust boundaries, data/perf, architecture-altitude, ponytail-simplicity, tests/failure paths), adversarial verification, dedup, ranked file:line report. Use proactively when asked to review, check, or assess a diff, branch, or PR — and after any non-trivial implementation, before the PR. Also on "review-swarm", "swarm review", "deep review". Not for trivial diffs, not when asked to fix rather than review, not when security alone is the whole ask (built-in "security-review"), and not when "/code-review" is named explicitly.

up-to-date

Preflight sync + situational brief before repo work: fetch origin, ahead/behind divergence, recent commits, open PRs/issues, dirty-state warnings, one recommended first action. Read-only — never pulls or rewrites a dirty tree without explicit OK. Use proactively before starting work in any repo with a remote, especially shared or long-untouched ones. Also on "up-to-date", "sync first", "pull latest", "catch me up on the repo". Not for repos without a remote or quick mid-task re-checks.