botfiles

Start a detached zellij session for a tracker-linked or natural-language task, and launch Codex or Claude as the default shell with an initial `$start-new-task ...` handoff prompt.

Initialize a new task folder with status.md, user_inputs/initial.md, user_inputs/input_artifacts/, task-progress-artifacts/, and task-progress-artifacts/scratchpad/. Supports GitHub issue URL inputs, Linear issue URL inputs, default Linear issue creation for trackerless starts, tracker-informed slugs, and machine/zellij metadata capture with optional live-session issue block sync.

Initialize a new task folder with status.md, user_inputs/initial.md, user_inputs/input_artifacts/, task-progress-artifacts/, and task-progress-artifacts/scratchpad/. Supports GitHub issue URL inputs, Linear issue URL inputs, default Linear issue creation for trackerless starts, tracker-informed slugs, and machine/zellij metadata capture with optional live-session issue block sync.

Rename Codex app thread labels and titles in local Codex state. Use when a user asks to rename, retitle, relabel, or make a Codex chat/thread easier to find in the Codex app sidebar or thread list, especially when the visible label still shows the first prompt such as "$start-new-task ZON-218".

Create compact Codex `/goal` plan files that require an implementation loop with Oracle GPT-5.5 Pro review between iterations. Use when the user wants a goal plan that explicitly sequences implementation, strongest-model Oracle review, rework from Oracle findings, repeated review until no material changes remain, and final verification.

Create a Linear ticket and immediately launch a detached zellij/Codex worker session for it. Use when the user asks to "create a ticket and start a session", "make a Linear task and $start-zellij-session-for-task", spin up a side worker for a new task, or repeat the common parent-orchestrator flow of drafting a scoped issue, linking it to a parent lane, and starting `$start-new-task ISSUE-ID`.

Create a Linear ticket and immediately launch a detached zellij/Codex worker session for it. Use when the user asks to "create a ticket and start a session", "make a Linear task and $start-zellij-session-for-task", spin up a side worker for a new task, or repeat the common parent-orchestrator flow of drafting a scoped issue, linking it to a parent lane, and starting `$start-new-task ISSUE-ID`.

Commit task-scoped work across personal_os task artifacts and any external codebases touched by the task. Use when the user asks to commit relevant files, save task work in git, commit personal_os context files, commit a task folder, or commit changes spanning personal_os plus another repo. Infer the active task and branches from context; stage only related paths; ask before committing an external repo when the branch or ownership is ambiguous.

Create self-contained HTML explainers that help a human understand the full context of a task, issue family, PR, architecture, or implementation plan. Use when the user asks for an HTML/webpage/visual explainer, wants background plus diagrams, asks whether related issues are worth taking up, needs an end-to-end mental model, or wants the artifact copied/opened on a local Mac from a remote VM.

Standardize end-of-task closeout when the user asks to finish, wrap up, merge-and-close, land, or otherwise cleanly close a tracked task or PR, including task-status sync, tracker notes/state changes, downstream heads-up checks, and local branch/worktree cleanup.

Use local read-only iMessage, WhatsApp, and LinkedIn channel access wrappers for raw message discovery, sample exports, and CRM provenance workflows.

Safely audit and clean macOS disk space with read-only baselines, approval-gated cleanup batches, Docker/Xcode/cache/log cleanup, verification, and task artifacts. Use when the user asks to free Mac disk space, clean Mac storage, reclaim disk space on macOS, investigate a near-full Mac, audit caches or developer artifacts, or create a repeatable Mac cleanup workflow.

Backup-first iPhone photo/video archiving to an external drive. Use when creating or verifying a Finder-browsable raw archive from an iPhone, iOS backup, or libimobiledevice backup before any phone media deletion or cleanup.

Run end-to-end deep research using OpenAI Responses API deep-research models plus Gemini Deep Research plus Exa Deep Research, then synthesize results into a comprehensive, easy-to-read, citation-backed Markdown report by default. Use for due diligence, investment research, market/competitor analysis, technical investigations, policy analysis, and any high-stakes question where cross-checking sources across multiple engines improves confidence; generate alternate formats only when the user explicitly requests them.

Guide a human through a task in an interactive, human-in-the-loop execution rhythm. Use when the user asks to be walked through something step by step, wants to confirm each action before receiving the next step, says they feel overwhelmed or need help focusing, or needs coordination across agent-run checks and human-run UI/admin/credential/risky actions.

Use the @steipete/oracle CLI to bundle a prompt plus the right files and get a second-model review (API or browser) for debugging, refactors, design checks, or cross-validation.

Compare, audit drift, and sync skills between botfiles-style visible skill trees (`claude/skills`, `codex/skills`) or hidden project-local skill trees (`.claude/skills`, `.codex/skills`), and sync local Codex skills against upstream skill repositories such as OpenAI curated skills or the Oracle skill with protected local forks. Use when the user asks to keep skills aligned across Claude and Codex, compare local skills to upstream, sync all or selected skills, or identify where drift exists.

Preview or send a bounded instruction into another known live zellij-backed coding-agent session outside Symphony.

Use repo-managed zellij helpers from ~/pro/botfiles to inspect, message, and launch parallel coding-agent task sessions.

Create a short Markdown goal-plan file for Codex CLI `/goal` usage. Use when the user asks to create, distill, or prepare a Codex goal plan from an existing task, accepted plan, status file, grill-me result, bug report, implementation plan, or when they want a crisp objective file for a long-running Codex goal.

Use Birdclaw for cache-first X/Twitter data workflows, especially followers/following, unfollowers, mutuals, non-mutual following, top followers, leadgen lists, local timeline/DM/search analysis, backups, or PR/code work in the Birdclaw repo. Prefer this skill over direct xurl whenever Birdclaw can answer from its local SQLite cache or explicit sync workflow; use xurl only for auth setup, unsupported endpoints, or intentional live API calls.

Use the xurl CLI for authenticated X/Twitter API work, including reading posts, searching, checking timelines and mentions, inspecting followers/following, posting, replying, quoting, sending DMs, uploading media, and using raw X API v2 endpoints. Use when the user asks to interact with X/Twitter through xurl or needs raw/unsupported X API workflows. Prefer the birdclaw skill first for cache-backed follower/following analysis, unfollower checks, mutuals, non-mutual following, top followers, and leadgen exports.

Edit any video by conversation. Transcribe, cut, color grade, generate overlay animations, burn subtitles — for talking heads, montages, tutorials, travel, interviews. No presets, no menus. Ask questions, confirm the plan, execute, iterate, persist. Production-correctness rules are hard; everything else is artistic freedom.

Create Excalidraw diagram JSON files that make visual arguments. Use when the user wants to visualize workflows, architectures, or concepts.

Gmail: Send, read, and manage email.

Resume an interrupted tracked task in Hermes by resolving the existing task home from a task slug, tracker ref, session name, task dir, or status file, then syncing task metadata to the current session before continuing.

Resolve another tracked live task/session and read its task metadata, transcript fallback, and zellij targeting details for non-Symphony Hermes orchestration.

Preview or send a bounded instruction into another known live zellij-backed coding-agent session outside Symphony.

Print rich task details for the current Hermes task context: status file path, primary tracker URL, machine name, session id, transcript path, zellij session/link, and a short task recap.

Save current task status, plan, or investigation notes to the project's task-tracking folder. Reconciles machine/zellij/tracker metadata and refreshes the current-task pointer for the active session.

Debug and fix detached Zellij launches for Codex tasks when session creation succeeds but the attached pane only shows a shell or no visible Codex UI.

Initialize a new tracked task folder with status.md, user_inputs/initial.md, user_inputs/input_artifacts/, task-progress-artifacts/, and task-progress-artifacts/scratchpad/. Supports tracker-aware slugs, canonical task roots, machine/session metadata capture, and immediate plan review.

Ralph Wiggum methodology for autonomous agentic development. Use when setting up a new Ralph task, preparing specs, or troubleshooting Ralph loops.

Ralph Wiggum methodology for autonomous agentic development. Use when setting up a new Ralph task, preparing specs, or troubleshooting Ralph loops.

Take over an interrupted tracked task from a previous Codex or Claude session by resolving the existing task home from a tracker ref, task slug, status file, or zellij session, then syncing the status file to the current session and reading the status plus transcript tail before continuing.

Take over an interrupted tracked task from a previous Codex or Claude session by resolving the existing task home from a tracker ref, task slug, status file, or zellij session, then syncing the status file to the current session and reading the status plus transcript tail before continuing.

Resolve another tracked live task/session and read its task metadata, transcript fallback, and zellij targeting details for non-Symphony orchestration.

Run a repeatable non-Symphony PR autoreview fix loop using current-head reviewer artifacts plus the repo-managed `pr-autoreview-loop` helper.

Choose a named Google Workspace account alias (`work`, `personal`, `columbia`) for `gws` and switch agents onto the matching exported credentials file. Use when a user wants Gmail or Drive read flows from a specific account or needs a practical multi-account workaround for googleworkspace/cli on this machine.

Google Drive: List, search, inspect, download, and export files with alias-first `gws-account` flows for `work`, `personal`, and `columbia`.