sibulus13
@sibulus13
Public Skills
tiktok-sync
by sibulus13
Post-review playbook for @chase.ride. For each reel in the Notion Motorcycle Content Bank that is "In progress" and was posted at least 24 h ago, match it to the TikTok post, write its metrics (views/likes/comments/saves/shares, watch time, % watched full, new followers), the per-second retention curve and the views-by-hour curve into the row, score it against the account's own history, and propose 3 simple tests for the next posts. Runs on a schedule (daily 21:30 PT) or by hand. Triggers "/tiktok-sync", "sync my reel metrics", "review my latest reel", "has my reel been posted".
fleet-status
by sibulus13
Real status check across every Foreman-managed dark-factory repo — what's done, what's actively dispatching, what's genuinely blocked vs. just queued, a duplicate sweep, and an audit of ideas/gaps that were diagnosed (Spec-Gap Ledger) but never actually queued for dispatch. Use whenever the user asks for a status check, "what's blocked", "what's coming up", or anything about Foreman's queue/fleet state.
wish-research
by sibulus13
Refine Cortex Deal Scout's research for a wishlist item with the owner's own findings — set budget/priority/category preferences, pin options they found, exclude brands or models, add notes — then re-rank and rewrite the top 3 picks and the Research sub-page. Use when the user says "/wish-research <item>", "update the research for my <item>", "I found this <product>, add it", "drop <brand> from my <item> picks", or wants to change what a wish is optimised for.
orchestrate
by sibulus13
Six named role-flows (feature / bugfix / arch-decision / security-review / go-to-market / hotfix) that sequence real agent personas with structural review gates, and hand multi-component fan-out to the fanout-design-build-audit workflow rather than re-deriving partition/execute logic ad hoc.
dark-factory
by sibulus13
The idea-to-product pipeline, built from ONE atomic unit, the cycle (Design → Design review → Implement → Review, size-scaled fix budget 1–3, every step traced under a cycle UUID). A build is intake/kill-gate, then a spec cycle, a contract cycle, parallel component cycles in their own worktrees, and an integration cycle. Entry modes (feature-add, spec-only, review-only, build-only, contract-only, integration-check, research, gap-log, consolidate) run any subset. It keeps docs/STATE.md and PHASE-LOG current so Catwalk shows status live. Full rules and rationale are in REFERENCE.md; the unit is specified in CYCLE.md.
grab-media
by sibulus13
Analyze a media URL (YouTube/Shorts, TikTok, IG, Reddit, X, Vimeo), then download it as video or audio into the content library. Reports title/duration/aspect/available quality BEFORE downloading, proposes a content-descriptive filename, defaults to the helpers folder, and verifies the merged output actually has sound. Triggers "download this video", "grab this short", "save this as a helper", "rip the audio from X".
miko-onboard
by sibulus13
Intake + configuration + go-live for the Miko (Pythia) chat widget on a NEW website. Use when adding Miko to a site — collect the tenant config, set up the tenant + CORS in pythia, author + ingest the site's corpus, embed a brand-matched widget on the public surface only, validate against the eval gate, and deploy. Complements /deploy-web (which ships the site; this configures the assistant). Triggers: "add Miko to <site>", "onboard <site> to the chatbot", "put the chat widget on <site>", "go live with Miko on a new website".
deploy-web
by sibulus13
Safe, preview-first production deploy loop for a web project (Vercel + Next.js/Vite/Expo). Use when shipping a site/app change end-to-end — identify the repo, verify setup + env, gate on a local build, deploy to a preview branch, check deployment health, and run an agentic end-to-end smoke test before promoting. Encodes the lessons from the si8/miko + cashcow/tarive deploys. Triggers: "deploy this", "ship it", "set up deployment for <repo>", "push a preview", "is this ready to deploy".
resume-research-hardening
by sibulus13
Session handoff contract from 2026-06-22. Resumes the anti-overfitting hardening of the backtesting pipeline. Invoke at the very start of the next session before doing anything else — reads current state, confirms what landed, and continues the two incomplete implementation tracks.
self-improve
by sibulus13
Run the self-improvement loop. Scans recent session patterns, clusters by category, filters by threshold, and proposes additions to CLAUDE.md/memory/skill files. Logs run to ~/.claude/improve/history.jsonl for dashboard display.
retrospect
by sibulus13
Run a friction retrospective. Use this skill when the user invokes /retrospect, asks to review session friction, wants to improve Claude Code configuration based on past sessions, or wants to add allow rules and update CLAUDE.md based on accumulated friction reports.
caveman
by sibulus13
Ultra-concise response style — one sentence max, direct answers
qa
by sibulus13
Validate an implementation against its spec acceptance criteria. Outputs per-item verdict (MET/PARTIAL/FAILED) plus regression risk analysis. Use after any Implementer agent completes.
brief
by sibulus13
Generate a role-specific context package for a downstream agent. Takes role + project + milestone. Strips irrelevant context and produces a minimum viable briefing that can be pasted directly into an agent prompt.