dotfiles

Dead-code forensics and spec cross-reference. Finds unreachable, orphaned, missing, or dormant code and classifies it DEAD, ZOMBIE, GHOST, or DORMANT. Use when the user says "find dead code", "what's unused", "what can I delete", "stale specs", "spec drift", "orphaned implementations", "find zombie code", or asks what's wired up vs sitting unused. Do NOT use for code-quality review (/age) or NIH/reinvented-wheel detection (/nih-audit).

Build lasting understanding of an unfamiliar codebase via a four-pillar model (Building Blocks, Entry Points, Infrastructure, Egress) plus an adaptive Socratic quiz, using tilth and Context7. Use when the user says "help me understand this codebase", "grok this repo", "onboard me", "learn this project", "memorize this codebase", "study this code", "walk me through this code", or "quiz me on this repo". Do NOT use for single-file scripts, repos under 500 LOC, or editing tasks — understanding only.

Map a concept end-to-end through a layered codebase — find the entry point, follow callers/callees layer by layer, and cross-check the impact radius. Use when the user says "trace this through", "map the X flow", "blast radius for Y", "what touches Z", "find the entry point for", "what's affected by (tilth), filesystem search (/scout), or dead-code detection (/ghostbuster).

Query coding-agent session logs (Claude, Codex, opencode) via DuckDB for usage analytics, tool patterns, error forensics, and routing decisions. Use when the user says "session analytics", "query my logs", "tool usage", "how often do I use", "check my sessions", "analyze my usage", or asks about tool/agent/skill behavior across sessions. Do NOT use for debugging current code, reading a single transcript, or questions about Claude's capabilities.

Render or apply a "routine env" — a named bundle of Claude Code plugins + MCP servers + skill packs defined in `envs/<name>.yaml` — into a setup-script body for Claude Cloud (Claude Code on the web), or apply it to the local machine. Use when the user says "set up a routine env", "provision my Claude Cloud environment", "install my cheese env", "render the setup script for <env>", "update my env", "scaffold a new env", or invokes "/routine-env". Do NOT use for cloud cron/scheduled agents that open PRs (that is `routine-scaffold`) or for a personal CLAUDE.md preferences overlay in a contributed repo (that is `claude-local`) — routine-env is about plugin/MCP/skill provisioning only.

Verify a spec's implementation against its requirements using Serena structural analysis, build verification, and test coverage. Use when the user says "verify the spec", "check spec implementation", "does this match the spec", "spec coverage", "verify acceptance criteria", or invokes /spec-verify; also after /cure to validate the result. Do NOT use for writing code or for general code review — use /age.

Detect and fix AI-generated code anti-patterns ("slop") across Rust, Python, TypeScript, Go, and Shell. Use whenever you generate or edit code, when the user says "de-slop", "clean up AI code", "remove AI slop", or during /simplify and /cook flows. Also trigger proactively as a pre-commit checklist on AI-written changes. Do NOT use for correctness or bug review — use /age or /code-review.

Detect and fix weak test assertions that pass even when code is broken, across Rust, Python, TypeScript, Go, and Shell. Use whenever you write or review tests, when the user says "strengthen assertions" or "fix weak tests", or during /cook, /press, and /simplify flows. Trigger proactively on test generation and review.

Create distinctive, production-grade terminal UIs and full-screen interactive CLI tools in Rust (ratatui/crossterm) or Python (Textual/Rich). Use when the user says "build a dashboard", "terminal UI", "interactive CLI", "TUI app", "system monitor", "log viewer", "file manager", or invokes /tui-design.

Locate a git worktree under ~/Dev by branch, slug, repo, touched path, open PR, or staleness, returning its path + status so you can cd / resume / inspect it. Use when asked "where's the worktree for X", "find the worktree touching <path>", "which worktree has the open PR for Y", "find stale worktrees", or when /worktree-find is invoked. Read-only — it locates, it does not remove.

Create or resume an isolated git worktree for a Claude Code task, keeping main clean. Use when asked to "create a worktree", "resume a worktree", set up an isolated branch for a task, or when /worktree is invoked. Requires a task slug.

Triage WARN/DIRTY worktrees into keep / archive / remove / stash recommendations, fanning out one read-only haiku content-digest sub-agent per worktree so the verdict rests on real contents (unique commits, uncommitted diff, untracked files), not just metadata. Use when asked to "triage worktrees", "what's in these stale worktrees". Recommends only — never removes, commits, or stashes anything itself.

Run git and GitHub operations inside a worktree you're not currently in, without tripping Claude Code's Seatbelt safety heuristics (compound cd+git, heredoc PR bodies). Use whenever you need to commit, push, or create PRs in another worktree — especially from orchestrators — or when about to write "cd <path> && git" or a heredoc gh PR body.

Interactive design verification via dependency-graph traversal (replaces /notebook) — point it at a module, spec, or PR. Use when reviewing large modules, verifying agent output, or auditing design, or when the user says "review this module", "verify the design", "is this the right architecture", "check this code against the spec", "what does this module actually do", or invokes /xray.

Reconstruct what a past coding-agent session was doing so you can resume it — goal, files touched, last verified state, and the next step — by querying the session logs. Use when the user says "what was I working on", "recover that session", "reconstruct where I left off", "resume my last session", "what did that session change", "rebuild context from logs", or invokes /work-recovery. Report-only — it never scores or judges. Do NOT use for usage scoring (that is /skill-improver, /tool-efficiency, /prompt-analytics) or one-off interactive log queries (that is /session-analytics).

Audit and improve an existing agent or skill definition. Use when the user says "improve this skill", "audit this agent", "optimize this agent", "review agent definition", "fix trigger rate", "skill not activating", or invokes /skill-improver with a path; also when an agent produces poor results and needs prompt tuning, or a skill isn't triggering reliably. Do NOT use for creating new skills from scratch — use /skill-creator for that.

Author, review, land, and register a Claude Code cloud routine end-to-end — a scheduled or event-triggered cloud agent that opens PRs/issues a human disposes. Use when the user says "make a routine", "schedule a cloud agent", "scaffold a watcher", "set up a recurring automation", "/routine-scaffold", or points at a recipe (drift-watcher, stale-docs, changelog-draft, repo-brief, flaky-test-triage, pr-review-nag). Drives the `coder` and `reviewer` phase agents and hands registration to the cloud-routine registrar (`RemoteTrigger`). Do NOT use for local scheduled tasks (the `create_scheduled_task` / plugin-`schedule` kind) or for one-off code changes (that is `/cook`).

Distill the user's global ~/.claude/CLAUDE.md into a gitignored CLAUDE.local.md for repos they contribute to but don't own — keeping only repo-relevant instructions and dropping personal flair. Use when the user says "set up CLAUDE.local", "scaffold local claude config", "drop my preferences in this repo", "I'm contributing and want my preferences applied", "claude-local.md", or invokes /claude-local. Also use proactively when they open an unfamiliar repo and want to start contributing.

Guardrail against git commands that read file contents outside the Read tool — especially `git show <ref>:<path>` and `git cat-file`, which bypass file-access controls. Triggers when constructing git commands with colon ref:path syntax or cat-file to read another branch or commit. Do NOT use for normal git operations like status, log, diff, commit, or push.

Diagnose and self-heal harness-config drift between live files (~/.claude, ~/.codex, opencode, Cursor, Copilot) and what `ap` renders from the dotfiles registries. Use when the user says "harness doctor", "check my harness config", "settings drifted", "why is this hook firing twice", or asks to audit agent config. In Codex, invoke via `$harness-doctor` or `/skills`, not `/harness-doctor`. Do NOT use for general code review (/age), single-file permission cleanup (/settings-clean), or app-level debugging.

Fan sub-agents over open GitHub issues and triage each into a verdict, grounding adversarially against code intelligence, the repo wiki, and external research. Use when the user says "triage the issues", "/rennet", "go through the open issues", "what's still open", "sweep the backlog". Do NOT use to FILE new issues (that's /harness-doctor) or to fix/implement them (that's /cook).

Clean a bloated .claude/settings.local.json by removing redundant, stale, and junk permission entries and ensuring hook-redirected skills are allowed. Use when the user says "clean settings", "prune settings", "settings cleanup", or invokes /settings-clean; also proactively when settings.local.json exceeds ~30 entries. Only touches settings.local.json (gitignored), never settings.json (committed).

Run Python test code in an isolated .claude/testing sandbox (via sub-agent) without polluting the main context, reporting only pass/fail counts and assertion details. Use when the user says "run a quick test", "verify this works", "sanity check", "test this snippet", or invokes /test-sandbox; supports --sweep to clean stale test files. Do NOT use for the project's real test suite — this is for quick snippets.

Diagnose and fix library version mismatches, dependency conflicts, and build file inheritance issues. Use when the user says "fix versions", "version mismatch", "dependency conflict", "why won't this build", or "update dependencies", and proactively when a build fails on version constraints — the fix is almost always a version bump, not restructuring the build. Use this before rewriting a build config or bypassing workspace inheritance.

Curate this repo's hallouminate wiki (.hallouminate/wiki/, the repo:dotfiles:wiki corpus) — add or update architecture pages, per-harness docs, and gotchas. Use when the user says "update the wiki", "document this in the wiki", "refresh the harness docs", "add a wiki page", "curate the wiki", "the wiki is stale", or invokes /wiki-curator. Also use at session end to write back a non-obvious decision or gotcha worth preserving. Grounds the existing wiki first, follows one-topic-per-file conventions, verifies every external doc URL before writing, and reindexes. Do NOT use for general code search (that is cheez-search) or for editing AGENTS.md command reference.

Use when the user wants deep, multi-session internalization of an unfamiliar codebase — not just an orientation, but the goal of being able to explain it on a whiteboard and answer "what breaks if I change X" without re-checking. Triggers on "grok this repo", "onboard me", "memorize this project", "lock this codebase into memory", "quiz me on this repo", "help me understand this codebase deeply". Runs a four-pillar march (Building Blocks → Entry Points → Infrastructure → Egress) plus a trace-one-request exercise and an adaptive Socratic quiz; persists artifacts to .cheese/grok/<repo>/. Read-only stance — DO NOT propose edits unless explicitly asked. For a quick single-session orientation ("tour this repo", "trace how X works"), use `/tour` instead. Especially tuned for TS/JS monorepos and Next.js / Express / NestJS / Fastify apps; methodology is stack-agnostic. Do NOT use for single-file scripts, repos under 500 LOC, or editing tasks.

Use when the user wants to interrogate an unfamiliar codebase with structured probes rather than a summary — invariants, data flow, error paths, hot paths, security surface. Triggers on "probe this", "what are the invariants here", "where does X flow", "find the risk in this code", "trace error paths", "what's on the hot path", "security audit this file/module", "what could go wrong here". Returns numbered findings with confidence + citations, never edits.

Use when the user wants a quick, single-session read-only tour of an unfamiliar codebase or a specific feature — orientation, not memorization. Triggers on "tour this repo", "give me a tour", "give me a read-only tour", "what does this project do", "map the architecture", "trace how X works", "where is Y implemented", "show me how this works", "explain this codebase quickly". Output is a layered summary (project purpose → module map → call graph for the pointed-at thing) with file:line citations. Stops after answering — never volunteers refactors. For deeper multi-session internalization with an adaptive quiz, use `/grok-codebase` instead.

Standard cargo workflow for build/test/lint cycles.

Analyze how prompts and skill routing behave across coding-agent sessions and produce calibrated recommendations — prompt-pattern analysis, routing accuracy, and knowledge gaps. Use when the user says "analyze my prompts", "prompt patterns", "is routing working", "which skill should have fired", "knowledge gaps", "what do I keep asking", or invokes /prompt-analytics. Do NOT use for auditing a single skill/agent definition (that is /skill-improver), tool/MCP efficiency (that is /tool-efficiency), or one-off interactive log queries (that is /session-analytics).

Audit how a tool, command, or MCP server is actually used across coding-agent sessions and produce calibrated recommendations — tool-vs-task fit, error forensics, fix recommendations, permission friction, MCP health, and token economics. Use when the user says "tool efficiency", "am I using X efficiently", "audit tool usage", "why does X keep failing", "how do I fix this error", "what should I change", "permission friction", "is this MCP worth it", "tool error rate", "fix recommendations", or invokes /tool-efficiency. Do NOT use for auditing a skill or agent definition (that is /skill-improver) or for one-off interactive log queries (that is /session-analytics).

Scan for custom code that duplicates well-supported libraries, then recommend migrations with effort estimates. Detects hand-rolled utilities, retry logic, validation, date handling, and DIY parsers. Use when the user mentions reinventing the wheel, asks "is there a library/crate for this", wants a build vs buy audit, says "what are we maintaining that we shouldn't be", or "should we just use lodash for this". Do NOT use for code-quality review (/age) or dead-code removal (/simplify or /ghostbuster).

Multi-source research orchestrator scoped to the todoist-flow plugin so profiles like `dots profile launch claude todo` can do fact checks without enabling cheese-flow. Spawns parallel fetch sub-agents for Context7 (library docs), Tavily (technical content), Serper (facts/SERP), and `gh` CLI (GitHub patterns). Fetchers write findings to scratch files; a single synthesis sub-agent reads them and returns a compact answer, keeping the caller's context clean. Use when the user is mid-task-triage and needs external facts before deciding whether to keep, reword, or extract a Todoist task, or invokes /todoist-flow:research directly. Do NOT use for single-source lookups (use Context7 directly for library docs alone). For coding sessions outside the todo profile, prefer cheese-flow's /briesearch instead.

Use when the user wants to draft, revise, critique, or shepherd a design document, RFC, ADR, tech spec, or design proposal. Triggers on "write a design doc", "draft an RFC", "ADR for X", "design proposal", "tech spec", "review my design doc", "tighten this design doc", "critique my draft". Drives a spine-first, anti-slop workflow where YOU author Context / Problem / Goals / Non-Goals / Alternatives / Risks and the AI is restricted to critique, expansion, and copyedit. Use even when the user says "help me write a design doc" — the help is procedural, not authorial.

Find Todoist tasks that are really reference material (not actionable) and export them as structured markdown for Dropbox or other storage. Use when the user says 'extract references', 'find non-actionable tasks', 'clean up reference tasks', 'move to dropbox', 'todoist to dropbox', 'separate reference from tasks', or when /todoist dashboard flags many noun-only or URL tasks.

Audit and restructure Todoist projects, labels, sections, and filters. Proposes changes with approval before executing. Use when the user says 'organize todoist', 'restructure projects', 'clean up projects', 'fix my todoist structure', 'audit labels', 'create filters', or when /todoist dashboard shows structural issues (too many projects, empty projects, missing filters).

Todoist dashboard and router. Shows inbox count, overdue count, project health, and routes to the right sub-skill. Use when the user says 'todoist', 'check my tasks', 'task overview', 'what's on my plate', 'show me my todos', or any general Todoist inquiry that doesn't clearly map to triage/organize/update/extract.

Interactive triage of overdue and stale Todoist tasks. Presents tasks in batches, user decides fate of each: complete, reschedule, delete, someday, or skip. Use when the user says 'triage', 'process overdue tasks', 'clean up todoist', 'deal with overdue', 'task triage', 'inbox zero', or when /todoist dashboard shows high overdue count.

Review active (non-overdue) Todoist tasks for completion, irrelevance, or rewording. Incremental hygiene, not bankruptcy. Use when the user says 'update tasks', 'review my tasks', 'task hygiene', 'check what I've done', 'mark tasks done', 'clean up active tasks', or when they want a project-by-project walkthrough of current work.