bob

Generate a self-contained, navigable explainer bundle for a feature of the current codebase — a sidebar of sub-concepts, detail pages, and linked animated diagrams grounded in the real code

Analyze a codebase and interactively generate an OKF (.knowledge/) bundle — asks clarifying questions, discovers packages and decisions, then writes a complete navigable knowledge catalog.

Analyzes code for unnecessary complexity, unjustified abstractions, and structural cleanup opportunities using first-principles engineering methodology

Finds bugs in existing code — nil dereferences, race conditions, resource leaks, logic errors, error handling gaps. Creates cleanup tasks for each finding.

Verifies that SPECS.md, NOTES.md, TESTS.md, BENCHMARKS.md, and documentation cross-reference cleanly against each other and against the actual code

Self-directed analyst that claims analysis tasks from a shared task list and writes findings (read-only)

Researches the codebase for a brainstorm task, evaluates multiple approaches, and writes findings to a brainstorm file

Self-directed challenger that claims completed analysis tasks and stress-tests them (read-only)

Implements one concrete set of tasks from the plan using TDD, then reports what changed

Creates a detailed TDD-first implementation plan from brainstorm findings and writes it to a plan file

Self-directed reviewer that claims completed tasks and reviews them incrementally

Specialized testing agent for running tests and quality checks

Autonomous brainstorming agent for workflow orchestration

EXECUTE phase coordinator that orchestrates implementation, task review, and code quality subagents

Adversarial deep code review — orchestrator spawns 8 specialist agents in parallel hunting spec drift, comment lies, memory hazards, concurrency bugs, contract gaps, code quality issues, and structural over-abstraction. Outputs severity-ranked findings to .bob/state/review.md with a routing recommendation.

Verify spec invariants and analyze codebase health — DISCOVER → AUDIT → [ANALYZE → SCORE] → REPORT → COMPLETE

Adversarial idea review — interactive intake followed by 8 specialist agents in parallel attacking assumptions, prior art, scope explosion, scale, operational burden, alternatives, user need, and reversibility. Outputs severity-ranked findings to .bob/state/idea-review.md with a routing recommendation.

Team-based code cleanup workflow - DISCOVER → PLAN → CLEANUP LOOP → COMMIT. Simplifies, removes complexity, fixes documentation. Never introduces new functionality.

Team-based code cleanup workflow - DISCOVER → PLAN → CLEANUP LOOP → COMMIT. Simplifies, removes complexity, fixes documentation. Never introduces new functionality.

Code review workflow orchestrator - REVIEW → FIX → TEST → COMMIT → MONITOR

Create or apply a spec-driven module structure — SPECS.md, NOTES.md, TESTS.md, BENCHMARKS.md, and // NOTE invariant on all .go files

Team-based codebase exploration with adversarial challenge - DISCOVER → ANALYZE → CHALLENGE → DOCUMENT

Team-based codebase exploration with adversarial challenge - DISCOVER → ANALYZE → CHALLENGE → DOCUMENT

Use when you have a spec or requirements for a multi-step task, before touching code

Break a large changeset into ordered, reviewable PRs — idempotent, call again to advance the stack after merges

Full development workflow orchestrator - INIT → WORKTREE → BRAINSTORM → PLAN → EXECUTE → TEST → REVIEW → COMPLETE

Team-based development workflow - INIT → WORKTREE → BRAINSTORM → PLAN → EXECUTE → TEST → REVIEW → COMPLETE

Team-based development workflow using experimental agent teams - INIT → WORKTREE → BRAINSTORM → PLAN → EXECUTE → REVIEW → COMPLETE

Generate a stylized, self-contained HTML report for any Go codebase element — functions, packages, interfaces, structs, handlers, CLI commands, and more

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Go coding guidelines for production-quality code — pool lifetimes, concurrency safety, numeric type boundaries, error handling, and test discipline

Multi-repo operational workflow — coordinate code changes across repos, build/deploy, and validate end-to-end. INIT → PLAN → CODE → OPERATE → TEST → COMPLETE

Write a blameless incident postmortem — gather timeline, root cause, impact, action items

Run a premortem — imagine a project has failed and work backwards to identify risks before they happen

Specialized planning agent for creating detailed implementation plans

Comprehensive multi-domain code reviewer that covers all review concerns in a single pass

Implements code changes following plans and specifications

Spec authority agent that scans SPECS.md/NOTES.md/TESTS.md/BENCHMARKS.md, answers invariant questions from teammates, and writes final spec doc updates

Go-specific pre-submit reviewer targeting patterns that survive generic review but cause problems in production — pool lifetimes, concurrency races, int64/int mismatches, early-stop correctness, spec drift, error handling, and test quality

Have a conversation with Codex about code — ask questions or instruct it to write code, looping until consensus.