dotfiles

Consult Fable (primary oracle) for expert second opinions; escalate to GPT-5.5-Pro only for extremely important or complex tasks (always paired with Fable). Use for complex decisions, architecture choices, debugging hard problems, or when user says "consult oracles", "ask the experts", or wants a second opinion.

Run iterative oracle + agent hardening loop on any artifact (designs, plans, beads, architecture) until findings converge to near-zero. Combines /swarm-oracle with /swarm-review in alternating rounds. Use for the full hardening cycle, not just a single oracle pass. For oracle-only, use /swarm-oracle. For bead-only hardening, use /swarm-beads-quality.

Run 2x oracle sessions (FOR + AGAINST stances) to validate design decisions, plans, or bead readiness. Default = two Fable subagents; escalate to PAL 2x GPT-Pro (always paired with Fable) for extremely important or complex validations. Use after design rounds, before implementation, or to challenge architecture decisions.

Act as a wise, effective teacher whose goal is to make the human deeply understand the work done in this session (a change, a bug fix, a feature, a design) — i.e. sync the human's mental model up to the agent's. Use when the user says "sync-human", "sync me up", "teach me this session", "make sure I understand", "walk me through what we did", "quiz me on this", or "I want to actually understand this PR/change", or otherwise wants Socratic, gated, incremental teaching with comprehension checks rather than a one-shot summary. Drives understanding at both high level (motivation, impact) and low level (business logic, edge cases) using a running checklist and quizzes.

MCP Agent Mail for multi-agent coordination. Use when agents need file locks, messaging, inboxes, or conflict prevention. Handles macro_start_session, file_reservation_paths, send_message, threading, pre-commit guards.

Retrieve API tokens, keys, and credentials Oystein has stored locally. Use whenever code, scripts, or shell commands need a secret value: GitHub tokens, Cloudflare, HubSpot, Slack, Zendesk, Jira, Sentry, Anthropic, Apify, Browserbase, Google OAuth, Huma. Use BEFORE searching shell history, session logs, dotfiles, or the filesystem — the canonical store is documented here and the values are reachable via two fish helpers. Also use when adding, rotating, or removing a credential.

Rebuild CHANGELOG.md files and release histories from git, tags, releases, and issue trackers. Use when writing changelogs, version timelines, or agent-facing project history summaries.

Find and fix concurrency bugs - deadlocks, races, livelocks, await-holding-lock, database locks, LD_PRELOAD init, swarm races. Use when processes hang, tests flake, or auditing concurrency.

Multi-perspective project analysis via NTM reasoning-mode agent swarm. Use when "modes of reasoning", "multi-perspective analysis", or "epistemological review" of any project.

Rank hot paths by CPU, memory, I/O, and contention; hand the optimization skill a scored target list. Use when: profile, flamegraph, hotspot, bottleneck, p95/p99, IOPS, fsync, "why is this slow".

Prepare project releases: run test suites, fix broken/obsolete tests, bump versions, build cross-platform binaries via GitHub Actions CI or dsr/rch fallback, create GitHub releases with checksums, verify installers. Use when: release, cut release, prepare release, version bump, pre-release, ship it, tag release, gh actions release, dsr release.

Tend NTM agent swarms: orchestrator loop, marching orders, autonomous recovery of stuck or rate-limited panes, review-only mode, and work coordination via Agent Mail, Beads, and BV.

Replace Tokio Rust stacks with Asupersync. Use when migrating tokio/axum/hyper/tonic apps, designing native Cx/region-based services, or debugging Asupersync internals.

Patch Claude Code's native binary to suppress two nag warnings: (1) "Auth conflict: Both a token and an API key are set" — for intentional dual-auth/proxy setups. (2) "Claude Code has switched from npm to native installer" — when keeping the npm install on purpose. Re-run after every Claude Code update (npm install or native). Use when user says "patch claude", "patch auth warning", "fix auth conflict", "remove native install warning", "suppress npm deprecation", or after updating claude.

Mine atuin history and shell logs to find repetitive workflows, build Rust CLI or bash automations. Use when analyzing command patterns or finding automation opportunities.

Beads Rust issue tracker (br). Use when tracking tasks, managing dependencies, finding ready work, or syncing issues to git via JSONL.

Graph-aware task triage with bv and br. Use when prioritizing work, finding bottlenecks, tracking dependencies, or managing local issues across projects.

Convert markdown plans into beads with dependencies using br CLI. Use when creating task graphs, polishing beads before implementation, or bridging planning to agent swarm execution.

Brenner Bot CLI for research sessions. Use when in your brenner_bot repo, managing hypotheses, searching corpus, or running multi-agent sessions.

Gemini 3.1 Pro code review swarm via NTM. Use when "gemini review swarm", "code review with gemini", or multi-round code auditing with Gemini agents.

Domain-parameterized codebase auditing (security, UX, performance, API, copy, CLI). Use when auditing code, assessing quality, finding issues, or pre-launch review.

Handle blocked destructive commands. Use when dcg blocks rm -rf, git reset --hard, DROP DATABASE, kubectl delete, or when configuring agent safety guardrails.

Build showcase-grade FrankenTUI screens. Use when working in ftui-demo-showcase, diagnosing TUI issues with doctor_frankentui, polishing TUI UX, or replacing placeholder interfaces.

Update dependencies to latest stable versions. Use when upgrading libraries, updating Cargo.toml, pyproject.toml, package.json, go.mod, Gemfile, or modernizing dependencies.

Find stubs, mocks, placeholders, TODOs, and fake code in a project. Use when "find mocks", "find stubs", "find placeholders", "check for fake code", or auditing for incomplete code.

Operationalize expert methods into corpus, quote bank, triangulated kernel, operator library, and validators. Use when distilling a methodology or mining session history into executable rules.

Audit and clean shell PATH pollution in .bashrc, .zshrc, .zshenv, .profile. Use when PATH has junk, wrong binary resolves, temp dirs in PATH, or duplicates.

Assess project status against README/plan vision. Use when "where are we", "reality check", "what's missing", "are we on track", "gap analysis", or "does this actually work".

Restore machine responsiveness via safe, selective process cleanup. Use when system unresponsive, high CPU/load average, IO pressure, filesystem cache bloat, memory pressure from btrfs/ext4, stuck tests, competing cargo builds, confused agents in loops, swap thrashing, disk full, systemd-oomd kills, or tmux/zellij session sprawl.

Design and implement metamorphic testing for systems with the oracle problem. Use when: testing ML models, scientific computing, compilers, search engines, databases, graphics pipelines, or any system where correct output is unknown but input-output relationships are predictable. Metamorphic relations, property-based testing, MR taxonomy, oracle-free verification.

Manage WezTerm mux servers for AI agent swarms on high-RAM servers (512GB+). Use when wezterm mux unresponsive, agent sessions need rescue, tuning scrollback, or configuring persistent remote sessions.

Search X (Twitter) archives for insights, conversations, and knowledge. Use when mining tweets, DMs, Grok chats, likes, or extracting information from X data exports.

Convert implementation plans into actionable beads with proper structure, acceptance criteria, dependencies, and test requirements. Use after completing an implementation plan and before starting development.

Generate images via OpenAI's gpt-image-2 (the SOTA OpenAI image model as of 2026). Use when the user wants to create, generate, render, or produce an image with OpenAI / GPT (e.g. "make me an image of...", "generate a logo with gpt-image", "use openai image gen", "openai image 2"). Wraps the v1/images/generations API in a small bash script.

Generate images via Google Gemini Nano Banana Pro (gemini-3-pro-image-preview), the SOTA Google image model as of 2026. Use when the user wants to create, generate, render, or produce an image with Google / Gemini (e.g. "make me an image with gemini", "use nano banana", "nano banana pro", "google image gen", "imagen alternative"). Wraps the v1beta generateContent REST API in a small bash script. Supports up to 14 reference images for blending and 4K output.

Get a second opinion on an image by asking the OpenAI Codex CLI to look at it. Use when the user wants Codex (GPT) to analyze an image they (or you) just looked at, asks for "a second opinion on this image", says "have codex look at it", or wants cross-model verification of a screenshot, photo, diagram, or any image file.

Spin up a team of parallel Opus agents to research, design, review, plan, or create beads. Generic multi-agent orchestration for any task that benefits from parallel exploration. Use when user says "spin up N agents", "research swarm", "design swarm", "review with N agents", or "plan with agents".

Run the full bead quality pipeline — review, oracle validation, hardening, and final verification. Iteratively improves beads until they are implementation-ready with zero issues. Use after creating beads or when beads need quality improvement.

Playbook for large-scale AI-assisted feature development. Covers 5 meta-patterns (two-track prioritization, escalating agent teams, oracle integration, convergence detection, fresh-eyes principle) plus full pipeline phases. Use when orchestrating a major feature from UX problem through implementation-ready beads.

Run 10 parallel review agents, each with a different quality lens (correctness, safety, perf, a11y, etc.), against plans, beads, code, or architecture decisions. Use when a thorough multi-perspective review is needed before shipping.