sirbepy_claude_config

Triggers on /clockify-reconciliator <project> only. Adds descriptions to description-less Clockify entries for a configured project, splitting large blocks into 1-3h chunks using git commits from configured repos.

Triggers on /work-recap only. Dispatches to a named recap variant (e.g. zirtue weekly, zirtue daily). Each variant lives in a subfolder under this skill.

Triggers on /schedule-once only. Schedules a SINGLE future run of a Claude prompt (or a raw shell command) on THIS machine via Windows Task Scheduler, as a self-deleting one-time task. The local, one-shot counterpart to the cloud /schedule and the recurring /cron-run. PC must be on and logged in at fire time; Claude Code itself need not be open. Also supports list and cancel.

Triggers on /create-pr only. Drafts a human-light PR for the current branch, scales the body to the diff, suggests visuals, previews locally, and creates it on approval.

Triggers on /autopilot only. Dev is AFK and wants maximum autonomous progress with heavy but purposeful token use. Never block - delegate aggressively to subagents to keep main context lean, resolve real judgment calls via a BOUNDED /iterate-it (capped per run), auto-answer any nested skill's question instead of hanging, log decisions, park true hard-stops, and grind the task to a verified finish.

Triggers on /batch-todos only. Dedupes ai_todos, classifies survivors as EASY (auto-execute) or HARD (dev picks), shows dry-run confirmation, batches all EASY todos, then surfaces the HARD queue.

Triggers on /bepy-project-setup-roblox only. Scaffolds a new Roblox/Luau project from the sirbepy_roblox/templates/default/ skeleton, or fills missing files in an existing one. Idempotent - never clobbers without asking. Verifies with aftman install + wally install + rojo build.

Triggers on /close only. Session retrospective, code-health review, persist, close terminal.

Triggers on /create-todo only. Writes a single ai_todo mid-session without waiting for /close - for a "note this for later" moment, a deferred fix, or a skill-improvement observation.

Triggers on /add-git-account only. Onboards a GitHub account for a project AND wires that project's commit identity + push auth + gh CLI to it, scoped by remote URL, without touching the global default.

Use before ANY creative or feature work - creating features, components, adding functionality, modifying behavior, or designing a change. Local lightweight replacement for superpowers:brainstorming. Explores context, asks only for genuinely unknowable facts, picks the approach internally, then implements. No approval gates, no spec hand-off, no plan sign-off.

Triggers on /commit and its subcommands (v, bump, onlybump, onlyv, push, pushbump, pushnbump) to commit changes.

Triggers on /character-creator <game-or-char> only. Game-centric by default - builds multiple characters from one game in a single batch (one icon source pass + one sound source pass shared across chars). For one-off single chars use `/character-creator <char> from <game>`. Outputs to <app-data>/characters/<char-id>/. Aborts cleanly when nothing usable is found. Never auto-triggers.

Use when you need to start a LONG-LIVED dev server / watcher for the current project (e.g. npm run dev, vite, next dev, flutter run, a backend that stays running). Routes it through the local server_supervisor app so it is visible in Joe's dashboard, owned (no orphans), and centrally logged - auto-registering it if needed. Do NOT use for one-off commands that exit (tests, builds, git, scripts); run those normally. Falls back to a normal shell run if the supervisor is not reachable.

Triggers on /ios-run only. Builds the current Flutter project on the dev's Mac over SSH and runs it on a connected iPhone, or builds a TestFlight .ipa.

Triggers on /code-check. Structural code review - file splits, DRY, dead code. Writes findings to ai_todos. Callable standalone or from /close.

Use when the user asks how much context is left or remaining, how full the context window is, how much room is left before it fills, or invokes /context-left. Reports the current claude session's remaining context-window tokens + percent. Read-only and localhost-only: queries the context daemon if it is running, else self-computes from this session's transcript.

Query Linear tickets via the GraphQL API. Triggers on any mention of a Linear ticket ID (e.g. MOB-123, ENG-45), or when the user asks to search, list, or look up Linear issues, projects, or sprints. Read-only - never mutates data.

Use when Joe asks what to delete / how to free disk space. Repeatable, self-improving cleanup scan that auto-routes by OS - macOS uses macos.md, Windows uses windows.md. Advise-only (Claude never deletes). Records new safe-to-delete spots and never-touch zones in the matching platform file, only on explicit /disk-doctor invoke.

Triggers on /pickup only. Reads the next-session handoff written by /next-ai-prompt, runs the verify checklist, then starts tackling suggested next steps.

Triggers on /cron-run only. Schedules a queue of LOCAL overnight agents using Windows Task Scheduler that grind through plan files in docs/night_run/, one task per tick, with review subagents and side-branch failure isolation. PC must stay on and logged in; Claude Code does NOT need to be open. For remote (PC off) use /night-run.

Triggers on /rate-it-and-commit only. Rates a change with /rate-it, then commits if it scores at or above threshold. Below threshold, offers iterate-first or commit-anyway. Default threshold 8.

Triggers when a skill needs to leave a structured handoff prompt for the next AI session. Called by /night-run finisher, or invoked manually. Writes .for_bepy/NEXT_AI_PROMPT.md with caller attribution. Never auto-triggers on its own.

Triggers on /night-run only. Schedules a queue of REMOTE overnight agents via RemoteTrigger that grind through plan files in docs/night_run/. PC can be turned off. For local (PC must stay on) use /cron-run.

Triage open Sentry issues across zng-app, zng-admin, zng-biller. Buckets by ACT/WATCH/REVIEW/CLOSE with delta tracking from a local snapshot.

Triggers on /project-onboarding only. Generates a fixed set of onboarding docs in .for_bepy/notebooklm/ by detecting the project stack (Flutter template ships today; other stacks prompt to scaffold a template or abort) and fanning out one subagent per content file. Overwrites existing files. Pass --compare=<path> to add a 1-page structural diff against another repo. Aborts and asks if stack is ambiguous or unknown.

Triggers on /shortcut-create-ticket only. Files a new Shortcut story for the dev, mirroring defaults from his existing assigned tickets, then logs the result.

Triggers on /iterate-it only. Converges a hypothesis through TWO phases: Explore (up to 6 rounds x 1 sub, varying angles, exit on score >= threshold) then Polish (up to 3 rounds x 1 sub, exit on score >= floor). One sub per round, sequential. Cheaper and finer-grained than the old N-per-round panel design. Use when a one-shot /rate-it isn't enough and you want the panel to PROPOSE BETTER VERSIONS, not just judge.

Triggers on /init-claude-md only.

Triggers on /roblox-animation only. Two subcommands - `author <name>` generates a placeholder KeyframeSequence Luau module; `import <url-or-path>` converts a Roblox Marketplace URL or a Mixamo BVH file into the same Luau module shape. Output is editable in Studio's Animation Editor.

Triggers on /bepy-skill-creator only.

Triggers on /zirtue-release-backfill only. Discovers all Shortcut tickets where the dev authored a commit in zng-app or zng-admin, then for each one whose Release field is unset / Next release / TBD / Oneday / V1.x, figures out the real release by matching merge SHAs against git tags. Backfills Release plus any other missing field metadata (Skill Set, Product Area, Technical Area, Priority, Estimate), then optionally moves resolved tickets to Complete. Always reports first, applies only after explicit approval.

Triggers on /rate-it only. Brutally honest 1-10 rating with named score tiers. No sugar-coating. Solo by default; pass an integer N to spawn an N-subagent panel synthesized to one verdict (use when stakes are high - panel mode costs ~5-6x tokens). Auto-detects if web research is needed; supports --research and --dont-research flags.

Triggers on /favicon only.

Triggers on /bepy-project-setup-web only. Web/HTML project standardization flow - runs favicon, meta-tags, PWA, GitHub Pages, styleguide, deploy workflow, etc. For Roblox/Luau projects use /bepy-project-setup-roblox instead.

Triggers on /fix-studio-content-folder only. Repoints HKCU Roblox Studio ContentFolder at a live Versions/version-<hash>/content/ folder when run-in-roblox panics with "system cannot find the file specified" after Roblox auto-prunes old install dirs.

Use when deleting files or directories. Picks the right tool per platform - PowerShell on Windows, Bash rm on Unix/Mac.

Triggers on /pwa only.

Triggers on /explain-commit only. Writes a structured deep-dive (TL;DR, context, old behavior, why it failed, the fix, trade-offs) of a git commit to .for_bepy/commits_explained/<id>.md. Use when the dev wants to understand or document why a commit was made.

Triggers on /github-board (or /gb) commands. Manages project work via GitHub Issues, creating epics, breaking them into tasks, planning conversationally, and implementing with a strict planning gate. Uses GitHub MCP to read/write issues. Use this skill whenever the user types /github-board, /gb, mentions epics, project planning, ticket management, or wants to track and implement features across their projects. Also triggers when the user says "add a ticket", "plan this feature", "what's on my board", "break this down", or references GitHub Issues in a project management context.