hex

Use when asked to audit a repository, assess codebase health, find what's wrong with a codebase, identify technical debt, or produce a prioritized improvement plan for an existing project. Also use when a repo feels untrustworthy — red CI, stale docs, mystery failures — and the user wants a grounded picture before investing, or says "upgrade this project", "what should we fix here", "how bad is this code".

Upgrade hex to the latest version from hex

Idempotently updates and restructures a repository's documentation — agent instruction files (AGENTS.md / CLAUDE.md, root and nested) and human docs (README, docs/) — converging them to a router-plus-topic-doc architecture and fixing doc↔code drift with code-cited evidence; re-running on an unchanged repo (prior run merged) produces zero diff. Use when asked to "clean up the docs", "restructure CLAUDE.md / AGENTS.md", "split this giant instruction file", "docs are stale", "fix doc drift", "set up agent docs for this repo", or after an audit confirms documentation findings worth fixing.

Delegate tasks to BOI v2. Covers TOML spec authoring, dispatch, context wiring, and monitoring. Use when delegating any non-trivial task — code changes, research, generation, or multi-step work.

Run the vibe-to-production pipeline on any Python project. Assesses code quality, generates characterization tests, prioritizes refactoring targets, and dispatches BOI specs for each phase. Use when the user says "vibe to prod", "harden this", "production-ready", "assess this project", "run assessment", or references the vibe-to-production playbook.

Validate hex agent structure and repair issues. Runs health checks against the expected directory layout, reports findings by severity, and auto-fixes safe issues. Also handles data migration from backup directories on first launch after bootstrap migration. Use when the user says "hex doctor", "check health", "fix my hex", "something is broken", or on first launch when .hex/migrate-from exists.

Hex memory system — save, search, and retrieve persistent memories across sessions. Provides FTS5 + optional vector hybrid search over indexed files.

Structure a decision when you're stuck. Turns "I need to think about X" into a clear options table with trade-offs and a recommendation. Saves the decision record so future sessions have the reasoning.

Adversarial analysis via conjecture and criticism. Spawns parallel agents to independently generate competing approaches, then each agent critiques all others. Consensus emerges from cross-criticism. Use when making recommendations with real alternatives, evaluating options, or when the user says "analyze", "evaluate", "criticize", "compare options", "what am I missing", or "poke holes".

Plan daily landings — concrete end-of-day outcomes prioritized by who you're blocking, what's blocking you, and your own deliverables. Also supports weekly target setting, mid-day check-ins, and open thread tracking. Use at the start of every work day after data pulls are complete. Also use when the user says "plan my day", "what should I land today", "daily landings", "set landings", "update landings", "how are my landings", "landing check-in", "set weekly landings", "weekly targets", or "weekly check-in".