amit-mizrachi
- 8 skills
- 0 followers
- 10 hours ago last updated
- ▌ Deep Review · amit-mizrachi bundleReview code with one deep single reviewer (Opus at high effort in Claude Code, the session model in Codex) using a bundle of public review best practices - superpowers code-reviewer format, 0-100 confidence scoring, Google eng-practices checklist. Use whenever the user asks to review changes, review code, do a code review, review this branch, review a PR's code, or runs /code-review. This replaces squad-review and agent-skills:code-review for every review.
- ▌ Fable Review · amit-mizrachi bundleReview code with one Fable subagent (the latest Fable model) running deep-review, wait for its report, then address the findings: verify each against the code, fix the confirmed ones, and re-review the fixes once. Use when the user runs /fable-review, optionally naming a PR number, a branch, or nothing (uncommitted work and the current branch).
- ▌ Address Review · amit-mizrachi bundleAddress the review comments on a GitHub pull request: read every unresolved inline thread, review body and PR comment, verify each against the code, apply the fixes that are needed, push, and reply on every comment saying what happened (Fixed in <sha> / Applied in <sha> / Skipped / Needs a human) with the reasoning. Use when the user runs /address-review, asks to address, answer or respond to PR review comments, or when night-sprint hands over external bot and human threads.
- ▌ Night Marathon · amit-mizrachi bundleTakes a feature from an idea to one pull request with no supervision between the stages, and asks the user nothing once it is invoked. It researches what the plan needs from the code and the allowed sources, publishes a short plan artifact with realistic UI mockups drawn from the repo's own design system and a pick list of the few decisions that matter, then builds the chosen plan overnight as one PR. Two modes: autonomous (the default - the recommended options are built without stopping) and review (only when the invocation says `mode: review` - the run stops at the plan artifact until the user pastes their picks into the conductor session). Use when the user says "night marathon", "feature e2e", "feature end to end", "research, plan and build this", "take this feature all the way", or wants a feature planned with mockups and then built while they are away.
- ▌ Product Research · amit-mizrachi bundleResearches a product question end to end with ONE approval from the user. It interviews the user, writes a research brief, and gets it approved; after that it asks nothing again. A background conductor turns the brief into a spec and research tickets (to-spec, to-tickets), runs a night-sprint in research mode that answers each question from the user's connectors and the web with cited sources, checks the evidence, and ends in one published artifact. Use when the user says "product research", "research this", "look into X for me", "competitive analysis", "market research", "what do customers say about X", or wants a question researched from Slack, Drive, email, CRM or the web without supervising it.
- ▌ Mywayfinder · amit-mizrachiWayfinder plus a published chart - the map's blocking graph as an interactive artifact, drawn when the map is charted and updated as the last step of resolving every ticket.
- ▌ Next Prompt · amit-mizrachiGenerate a continuation prompt for the next task and launch it as a new background agent (visible in the Claude Code Agents screen). Use when the user wants to hand off work to a fresh session, is running low on context, or wants to capture what to do next.
- ▌ Night Sprint · amit-mizrachi bundleDelivers a whole feature overnight through autonomous sessions run strictly one after another - a conductor session that writes no code but launches each ticket, revives stuck or dead sessions, and fires the reviews, plus one implementer session per ticket, all on ONE branch landing as ONE pull request. Gets or builds a ticket breakdown first (via to-spec and to-tickets), decides whether to review once at the end or at checkpoints, and optionally ends with a test session that boots the stack or runs evals. Use when the user says "night sprint", "sprint this feature", "build this overnight", "run this while I sleep", "ticket after ticket", or wants a feature taken end to end unattended in a single PR.