dotfiles

Coordinate a two-agent planner/reviewer loop across live Emacs agent sessions. Use when the user asks to orchestrate Claude/Fable and Codex, run paired model review, converge a plan through back-and-forth review, monitor agent sessions until completion, or make a multi-agent review process unattended.

Use when a LuLu firewall alert is open on this Mac and the user asks whether to allow or block it, or asks to inspect, check, audit, or review the connection a LuLu alert is prompting about. Covers explicit $review-lulu-alert requests and questions like "should I allow this?", "what is this process connecting to?", or "is this LuLu prompt safe?".

Dotfiles and Emacs package routing context. Use when working in ~/My Drive/dotfiles, editing emacs/config.org or emacs/extras, choosing between the dotfiles repo and an elpaca source clone, or applying dotfiles documentation and commit conventions.

Dotfiles and Emacs package routing context. Use when working in ~/My Drive/dotfiles, editing emacs/config.org or emacs/extras, choosing between the dotfiles repo and an elpaca source clone, or applying dotfiles documentation and commit conventions.

Audit a macOS .app before running it or granting permissions, especially when the user asks whether a Mac app is safe, suspicious, malware, from an unknown developer, or requests screen recording, accessibility, camera, microphone, input monitoring, or other sensitive permissions.

Scan for exposed API keys and leaked credentials, audit dependency vulnerabilities, review macOS hardening settings, and evaluate Claude Code configuration risks. Use when the user wants a security review, vulnerability scan, dependency audit, secrets check, or periodic security posture assessment.

Diagnose agent, tooling, configuration, or workflow failures before fixing. Use for /diagnose, why-did-this-happen questions, process friction, broken hooks/skills/scripts, or structural-remedy investigations; not for ordinary code bugs.

Diagnose agent, tooling, configuration, or workflow failures before fixing. Use for /diagnose, why-did-this-happen questions, process friction, broken hooks/skills/scripts, or structural-remedy investigations; not for ordinary code bugs.

Fetch recent tweets from a curated account list, triage them against its rubric, save an org digest, and update last-run state. Use twitter for direct lookups, twitter-discover for finding accounts, and twitter-vet for account scoring.

Discover high-value Twitter/X accounts in a niche through criteria, seeds, graph expansion, and vetting. Use for find accounts, who should I follow, curate a list, or twitter discovery; not for direct lookup, digests, or known-list vetting.

Fetch, read, summarize, or save specific Twitter/X tweets, threads, replies, profiles, timelines, follower lists, or searches. Use twitter-digest, twitter-discover, or twitter-vet for digests, account discovery, or account scoring.

Vet, score, or decide whether to add Twitter/X accounts to a curated list. Use when asked to vet a handle, score accounts, decide whether an account belongs on a list, or when twitter-digest/twitter-discover need shared account scoring.

Fetch recent tweets from a curated account list, triage them against its rubric, save an org digest, and update last-run state. Use twitter for direct lookups, twitter-discover for finding accounts, and twitter-vet for account scoring.

Discover high-value Twitter/X accounts in a niche through criteria, seeds, graph expansion, and vetting. Use for find accounts, who should I follow, curate a list, or twitter discovery; not for direct lookup, digests, or known-list vetting.

Fetch, read, summarize, or save specific Twitter/X tweets, threads, replies, profiles, timelines, follower lists, or searches. Use twitter-digest, twitter-discover, or twitter-vet for digests, account discovery, or account scoring.

Use when correctness depends on exercising a real running software workflow through its user-visible or runtime surface: interactive UI, browser, live Emacs, terminal display, async or network delivery, persistence, or scheduler behavior, especially when asked to try, reproduce, or verify it for real. Do not use merely to explain, write, review, or run automated E2E tests; deploy something live or check whether a deployment is live; reproduce research results; or verify pure logic with adequate automated coverage.

Use when correctness depends on exercising a real running software workflow through its user-visible or runtime surface: interactive UI, browser, live Emacs, terminal display, async or network delivery, persistence, or scheduler behavior, especially when asked to try, reproduce, or verify it for real. Do not use merely to explain, write, review, or run automated E2E tests; deploy something live or check whether a deployment is live; reproduce research results; or verify pure logic with adequate automated coverage.

End-of-session bookkeeping. Do NOT invoke this on your own initiative, and do NOT hand-write a session log yourself instead of invoking it — writing or editing any file under a project's logs/ directory is this skill's job alone. Run it only when the user explicitly asks (the user types /update-log or says update log, session log, close out, wrap up, or asks to save project progress) or when the deliberate end-of-session chain invokes it with --auto. Never trigger it proactively mid-session or at session end on your own judgment, even when the session clearly changed durable project state — wait for an explicit request or the --auto chain.

Process a list file strictly item by item, sequentially or with a bounded subagent pool. Use when the user needs no skipping, batching, or peeking ahead; not for lists that can be safely read all at once.

Batch-process org-roam TODOs autonomously. Acts on TODOs that can be completed without user input and records blockers for the rest, sorted by ease of unblock. Use when the user says "overnight todos", "run overnight", "process my todos", "todo batch", "act on my todos", or schedules a nightly run.

Rewrite prose to remove patterns that read as AI-generated — vocabulary tells (delve, tapestry, pivotal, underscore), copula avoidance (stands as, serves as, marks), negative parallelism, rule-of-three stacks, trailing -ing summary clauses, significance puffery, compulsive summaries, em-dash overuse, mechanical bolding, bullet-with-bold-header lists, render artefacts (oaicite, contentReference, turn0search0), and similar fingerprints. Use for /humanize, de-slop, removing AI tells, "sounds like ChatGPT/Claude/Gemini", anti-AI-detection cleanup, or as the final pass before publishing AI-assisted prose. Also invoked by other skills (e.g. wikipedia-article) as a last-mile cleanup phase.

Draft in Pablo's own voice anything he will send or publish as himself — email, Slack, forum/PR comments, tweets, Google Docs/Sheets comments, letters, reviews. Use whenever composing text that will go out under Pablo's name, even a one-line reply or a quick DM. It matches his private voice samples AND avoids AI-writing tells in a single composition pass, so there is no separate cleanup step afterwards. Do NOT use for text Pablo is not personally authoring (use `humanize` for that), nor for internal notes, commit messages, or code.

Rebuild and republish the CR-task master tracker artifact. Use when the user says "refresh cr tracker", "update the CR task tracker", "/refresh-cr-tracker", or wants the conceptual-reasoning task master list re-pulled from live reasoning-tasks data + GitHub issues and redeployed to its claude.ai artifact.

Diagnose a frozen, hung, beachballing, or unresponsive Emacs by sampling the process, reading server buffers when possible, and recommending recovery. Use when Emacs is stuck or not responding; do not use for ordinary Elisp errors while Emacs is responsive.

Proofread Markdown for spelling, grammar, typos, punctuation, and light style/clarity issues. Use for proofread, spell check, grammar check, typo review, or proofreading a Markdown file.

Proofread Markdown for spelling, grammar, typos, punctuation, and light style/clarity issues. Use for proofread, spell check, grammar check, typo review, or proofreading a Markdown file.

Use when writing, editing, reviewing, or testing Emacs Lisp code, including dotfiles extras packages, Elpaca rebuild/reload behavior, batch-test verification, docstring and error-message conventions, and transient menu suffix checks.

Use when writing, editing, reviewing, or testing Emacs Lisp code, including dotfiles extras packages, Elpaca rebuild/reload behavior, batch-test verification, docstring and error-message conventions, and transient menu suffix checks.

Release one of the user's Emacs packages or audit package release readiness. Use when the user says `/release-package <package>`, asks to bump/tag/create a GitHub release, wants release notes for a benthamite Emacs package, or runs `/release-package --audit`; do not use for dotfiles releases.

Design and build a reusable AI automation. Use for automate, create automation, turn this into a skill/command, new workflow, AI helper, or deciding whether a workflow should become a skill, directive, tool, or Emacs command.

Design and build a reusable AI automation. Use for automate, create automation, turn this into a skill/command, new workflow, AI helper, or deciding whether a workflow should become a skill, directive, tool, or Emacs command.

Use when the user asks Claude about the Codex App in-app browser handoff skill or why `$codex-app-browser` works only in Codex.

Use when the user explicitly invokes `$codex-app-browser`, asks to use the Codex App, in-app browser, in-built browser, Browser Use, or to open/navigate/click/test/screenshot something in the Codex app; also use when the user says they switched or resumed a Codex CLI/Emacs session in the Codex App to get browser access.

Review the current session when a Stop/session-end hook asks for lesson capture, then write candidate reusable lessons and promotion metadata to the central dotfiles .agent-learnings/inbox without implementing them. Use when a hook prompt says to capture session lessons, review current-session learnings, populate the agent-learning inbox, or preserve reusable agent self-improvement ideas.

Relocate Codex session logs by rewriting embedded project paths so resume and log lookup associate a session with the current project. Use for move session log, import log, or project rename session cleanup.

Use when creating or editing an org-mode note file. Applies note-structure and prose-formatting conventions for private or personal Org notes, regardless of directory.

Optimize project agent instruction files. Use for optimize/improve/clean up/review Claude.md or making persistent CLAUDE.md or AGENTS.md guidance more concise, concrete, and reliable; not for ordinary docs or code reviews.

Diagnose a frozen, hung, beachballing, or unresponsive Emacs by sampling the process, reading server buffers when possible, and recommending recovery. Use when Emacs is stuck or not responding; do not use for ordinary Elisp errors while Emacs is responsive.

Close the loop after agent pushes by waiting for GitHub Actions, inspecting failed logs, fixing push-triggered CI failures, and repeating until the pushed commit is green or a real blocker is identified. Use whenever an agent pushes commits or the user asks to check CI after a push.

Use when adding works to Pablo's bibliography, creating BibTeX/BibLaTeX entries, adding DOI/ISBN/URL references, resolving missing citekeys, or preparing notes that cite works not yet in the configured bibliography files.