Every DHF step, every owning skill.

Reads repo guidance, recovers current phase, classifies the lifecycle stage, selects the next skill/helper route, and defines evidence gates.

Example repo-specific lifecycle adapter. Takes over when a known repository needs domain paths, local commands, fixtures, or deployment boundaries.

Analyzes complex requirements, architecture context, task decomposition, and implementation risks before code changes.

Converts requirements or principles into an executable development package with plans, safety net, contracts, acceptance tests, and owner handoffs.

Turns fuzzy implementation intent into a backlog-ready issue, ticket, or executable spec while preserving DHF source-of-truth, lane, and verification boundaries.

Owns the architecture alignment checkpoint when durable sources conflict: compare product, process, components, data ownership, lane, security, slices, and handoff state before edits.

Reviews architecture, data flow, edge cases, test coverage, performance, and operational risks; can use structured planning context when gbrain is configured.

Pressure-tests product framing, scope, user value, strategy, and whether the plan should expand, hold, or reduce scope.

Runs founder-style or builder-style office hours to expose demand reality, narrow the wedge, and clarify the product idea; can use gbrain preflight when configured.

Reviews UX structure, visual direction, responsive behavior, and design acceptance criteria; gbrain context is specialist planning input, not a substitute for repo state.

Orders the execution workflow across Karpathy-style thinking, planning, TDD, and verification for complex work.

Use throwaway prototypes only to answer one named data model, state machine, module interface, or UI question, then delete the prototype or capture the durable decision.

Enforces a RED -> GREEN -> REFACTOR loop for feature work, bug fixes, refactors, and core business logic.

Starts from business-facing acceptance scenarios, then drives implementation through executable specifications and production-like checks.

For bugs, wrong output, broken UI flows, or unclear root cause, establish a runnable feedback loop before patching from inspection.

Runs the delivery verification pass across build, static checks, tests, coverage, and security scan where applicable.

Requires completion evidence with command, exit code, key output, timestamp, demand-matched gate selection, effective feedback check, and conversion health status when local evidence exists.

Performs pre-landing diff review for structural risks, trust boundaries, conditional side effects, and merge readiness.

Reviews staged, unstaged, or PR changes for correctness, security, maintainability, and project standards.

Runs parallel read-only review agents across behavioral, security, performance, reliability, and test coverage risks.

Runs infrastructure-first security and privacy review for auth, tokens, secrets, PII, approvals, audit trails, and public/private boundaries.

Checks input validation, authentication, authorization, secret leakage, injection, and error disclosure risks.

Runs systematic browser QA, screenshots, console/network checks, responsive checks, and iterative bug verification.

Runs report-only browser QA when the user wants findings, screenshots, and risk notes without automatic fixes.

Only explicit committee, expert panel, reviewer/worker, rating loop, or target-score requests route here; ordinary reviews keep their specialist route.

Creates a three-domain expert committee, collects ratings and findings, gives a revision brief to a worker, and repeats until the target rating or a blocker is reached.

Runs the opinionated ship workflow: detect base branch, test, review diff, update release metadata, commit, push, and prepare PR.

Merges and deploys after ship, waits for CI/deploy completion, and verifies production health.

Performs post-deploy canary monitoring for console errors, performance regressions, page failures, and screenshot drift.

Creates engineering retrospectives from commit history, work patterns, quality metrics, and team-aware summaries.

Reviews, searches, prunes, and exports project learnings so future sessions reuse operational knowledge.

Tunes repeated planning questions, developer profile signals, and question preferences; this is operator preference state, not repo lifecycle state.

Performs post-ship documentation updates across README, architecture docs, changelog voice, TODOs, and version notes.

Updates docs, codemaps, README, guides, paths, and examples after feature, API, or architecture changes.

Evaluates whether a skill should exist, how it should route, what evals it needs, and whether it improves behavior over baseline.

Builds self-contained HTML diagrams, flowcharts, architecture views, and visual explanations for complex technical concepts.