Deep-expertise parallel-execution orchestrator for closing a whole EPIC — a parent task with many child tasks — at speed without dropping the verification bar or violating the "nothing is done until the human says so" rule. Triggers whenever a user wants to drive an epic (or any large multi-task backlog cluster) to completion in parallel: "let's knock out epic X", "parallelize epic X", "review all the tasks in this epic and start implementing", "we're accumulating tasks faster than we implement them — speed this up", "orchestrate this epic across worktrees/subagents", or points at a named epic and asks to complete it. Also triggers when a user wants a repeatable, tested pattern for running many related code tasks concurrently in isolated worktrees with adversarial review before implementation. Use it for any epic you must review, de-conflict, implement, and verify in parallel — with evidence, isolation, and a human gate at the end.
Deep-expertise codebase cartographer and auditor. Reconstructs an unfamiliar codebase — especially a large multi-repo product line — from source alone, read-only, and produces onboarding maps, architecture deep-dives with diagrams, a severity-ranked issue register, and a system-level synthesis. Triggers whenever a user wants an unfamiliar codebase documented, mapped, explained, audited, reverse-engineered, or reconstructed; whenever they ask "what is this thing?", "how does data actually flow through this?", "what would bite me here?", "is this safe?", "where would this lose data?", "what does this actually ship?", or "onboard me to this repo" — even if the words "document" or "audit" are never used. Also triggers for auditing a data path, a credential path, or a durability claim in code you cannot run. Also triggers for INVENTORY questions about a codebase — "how many lines of code do we own?", "how much of this is actually ours vs upstream?", "how much of it is still maintained?", "how much of this is test
Deep-expertise startup and early-stage business plan analyst. Triggers whenever a user shares a business plan, pitch deck, executive summary, startup concept document, market analysis, or asks Claude to evaluate, critique, validate, stress-test, or improve any business proposal. Also triggers for questions like "is this viable?", "what am I missing?", "what would an investor say?", "how much would this cost to build?", or "who might partner with us?" — even if the word "business plan" is never used. If there is a document describing a product, service, or company and the user wants analytical feedback of any kind, use this skill.