iansteitz1-eng
UserOpen-source Claude Code Skills that replace clicking through vendor dashboards — Stripe, Cloudflare DNS, ElevenLabs, GitHub, and more. From InsyncTech, the company behind Vox Ordo.
Categories
Indexed Skills (25)
ffmpeg-audio
The prod audio convention made declarative — coerce any container (.m4a/.mp3/.webm/.wav) to mono 16 kHz signed-16-bit PCM WAV (`-ac 1 -ar 16000 -c:a pcm_s16le`), the input shape voiceprint, diarization, whisper, and training pipelines expect. Four ops (normalize · trim · concat · probe) plus a YAML reconcile mode. Idempotent, dry-run by default (nothing runs without --apply). Use when the user says "normalize this audio", "make a 16k wav", "extract an enroll clip", "convert to mono 16k", "prep audio for whisper/diarization", or "batch-convert these recordings".
question-economy
Near session end, when you're winding down, or on request, analyze the questions Claude asked you and your answers — then distill each answer into a standing default written to memory, so Claude asks fewer (only genuinely novel) questions over time. Tracks a per-session question-rate metric so the decline is measurable. Use when you say "wrap up", "winding down", "what did you learn about working with me", "question economy", or proactively at session close.
session-cabinet
Use when the user says "file my sessions", "session archive", "where are my old sessions", "make my Claude history searchable", "organize my Claude Code transcripts", "I can't find that session from last week", or "index my Claude sessions". Auto-files Claude Code session transcripts (~/.claude/projects/**/*.jsonl) into a human-navigable date-anchored archive with semantic SESSION.md cards, per-month _INDEX.md tables, and a master index. Idempotent — safe to run on a schedule. Free skill, stdlib only.
skill-manifest-gen
Generate cross-harness discovery manifests (openai.json · gemini.json · mcp.json + README) so an aria-skill is findable from OpenAI/Codex/Grok, Google Gemini, and MCP clients (VS Code, Cursor, Claude Desktop) — not just Claude Code. Derived deterministically from SKILL.md + the script's argparse, so manifests never drift. --all sweeps the repo, --check is a CI drift gate, --force overwrites. Use when Ian says "generate manifests", "MCP marketplace", "discoverable in VS Code/Cursor", "backfill manifests", or after adding a skill.
sprint-scaffold
Scaffold a new sprint folder with the Filing Cabinet spec_charter.md + Flowstate .claude/agents/pr-review.md + reference/sql/ subdirs. Use when starting a new sprint (e.g. "open sprint 021", "start a sprint for X", "new sprint folder"). Drops a working skeleton in ./sprints/NNN_slug/ with the load-bearing files pre-filled so every shipped item gets an audit-trail home. Validated over a two-terminal sprint that shipped 20+ items under this pattern.
deploy-verify
Post-restart proof that the SPECIFIC change you just shipped is the code now serving traffic — not just that the service is up. Runs data-driven assertions (health body, grep the live file for your wiring, import the new module in the service env, grep the SERVED HTTP asset, scan journal for errors since restart) and returns one PASS/FAIL verdict. Use after any deploy/restart, when Ian says "did it go live", "verify the deploy", "is my change live", "confirm it shipped", or to prove a fix in the running product (not code-trace).
safe-restart
Pre-restart multi-lane in-flight guard. Before restarting a shared Aria service (ac_gateway), checks whether ANY other Claude Code lane has uncommitted edits staged on the prod box — because a restart is one shared checkpoint that ships everyone's mid-flight work at once. Returns a GREEN/HOLD verdict (also gates on live voice sessions). Use when Ian says "is it clear to restart", "safe to restart", "anyone close to shipping", "who else is editing prod", "pre-restart check", "fleet check", or before ANY shared ac_gateway restart.
trace-write-target
Given a writer module/function on the server, determine which DATABASE and TABLE it actually writes to — accounting for sys.path injection and per-emitter DSN env-var selection. Answers "I queried the obvious pool and the table doesn't exist — where does this data actually land?" in one shot instead of several round-trips. Read-only. Use when Ian says "where does this write to", "which DB does X use", "the table doesn't exist", "where do the splats go", "trace this emitter", or before querying for rows a writer produced.
goal-to-plan
Turn a vague/ambiguous ask into an agent-ready plan — the "quality of the brief" discipline an AI-native team runs on. Emits a spec_charter (mission · scope · ONE acceptance criterion · risks) + a ready-to-dispatch lane brief (role · owned files · acceptance · report protocol), chaining the premortem + sprint-scaffold skills. Use when the user says "plan this", "turn this into a brief", "scope this out", "make this agent-ready", "what's the plan for X", "goal to plan", "/goal-to-plan", or before dispatching a lane / starting multi-step work from a fuzzy goal.
premortem
The DVE/GVE prosthetic premortem reflex — before writing, changing, debugging, or deploying code, briefly imagine how the decision could fail and turn that into an ARTIFACT (predicted failure modes → cheapest guard → predicted-vs-actual ledger), not just talk. Use before any risky change or deploy, or when the user says "premortem", "how could this fail", "what's the risk", "/premortem", "/dve". The operational core of the DVE Guidebook tool.
surgical-patch
Drift-safe editor for a hot, shared, possibly-remote file (e.g. prod main.py that's AHEAD of your local checkout). Applies a data-driven list of exact-match edits with three guarantees — idempotent (sentinel no-op), exact (every anchor must match a precise count or it aborts writing nothing), and reversible (timestamped backup + verify-or-rollback). Works locally or over ssh. Use when Ian says "patch the live file", "surgically edit main.py", "apply this change to prod without clobbering", "the server file is ahead of mine", or before any edit to a shared file multiple lanes touch.
video-ingest
Pull metadata + transcript from any YouTube, Instagram, X/Twitter, TikTok, Vimeo, or other yt-dlp-supported video URL so Claude can actually "watch" it. Use when the user pastes a video link and asks "watch this", "what's this video about", "summarize this", "let's talk about this video", or any time a URL points at video content. First tries subtitles (fast, no GPU). Optional `--whisper` fallback transcribes audio locally via faster-whisper.
aria-skill-candidates
Observer + Gatekeeper for the Aria Builders meta-loop. Scans Claude Code transcripts for recurring vendor-API patterns and ranks them by a 5-axis rubric (recurrence · YAML-shape · multi-step · public-value · sovereignty-cost) to decide which patterns should become new skills. Filters out vendors already covered by an existing skill. Use when the user says "what should I build next", "skill candidates", "where am I repeating myself", "scan transcripts", or weekly as a passive review.
aria-skill-template
Scaffolds a new aria-skills-pattern skill from a template. Generates README.md + SKILL.md + script.py with reconcile-loop stubs + config.yaml + requirements.txt — a fully-formed skill directory ready to fill in with vendor-specific logic. Multiplies contributor velocity. Use when the user says "scaffold a new skill", "create a skill for X", "new skill template", "I want to build a skill for <vendor>", or starting any new vendor integration.
aria-skill-test
Regression harness for the aria-skills repo. Runs every shipped skill in dry-run mode + asserts clean output (no traceback, no credential leaks, valid SKILL.md frontmatter, README.md present). Text report by default; --junit-xml flag for CI integration. Use when the user says "test the skills", "regression check", "before push", "run the test harness", or any time before opening a PR or pushing to main.
aria-status
Config-driven one-screen status of a server stack — systemd timers, healthz endpoints, disk usage, GPU memory/util (if nvidia-smi present), and optional custom shell-command probes. All probes run in parallel with short timeouts; graceful degradation if any single check fails. Read-only by design. Text output by default; --json for machine-readable, --brief for one-screen. Use at session start, before a deploy, or when the user says "status", "how are things", "any issues".
call-scrub
Pull and analyze your ElevenLabs Conversational-AI voice calls — list recent conversations, or scrub one transcript to surface every tool the agent called, the result it got, errors, and timing. The fast way to debug "why did that call misbehave" without replaying audio. BYOK (your own ElevenLabs API key); read-only; stdlib only. Use when the user says "scrub that call", "what happened on the call", "pull the call transcript", "debug the voice call", "what tools fired".
cloudflare-dns-deploy
Reconcile DNS records from a declared YAML catalog to Cloudflare via the API, and optionally run the companion nginx + certbot bootstrap for records pointing at your server. Idempotent. --dry-run by default, --apply hits the API, --server-setup also wires nginx + obtains Let's Encrypt cert. Use when the user says "deploy DNS", "add a subdomain", "stand up <X>.example.com", or before any subdomain launch.
doc-to-pdf
Render a Markdown or HTML document into a clean, branded, multi-page PDF via headless Chrome — the polished deliverable you'd hand a client, not a screenshot. Brand wrap (print-CSS + footer) is on by default; --no-brand for raw; --template for a custom shell. Never overwrites without --force. Use when the user says "make a PDF", "turn this into a PDF", "PDF this", "send them a PDF", "build a client doc", "export to PDF", "branded PDF".
github-repo-deploy
Reconcile a YAML catalog of GitHub repo metadata (description, homepage, topics) to the live GitHub account. Idempotent, --dry-run by default, --apply --prod for the live PATCH. Auth via `gh auth token` or GITHUB_TOKEN env. Refuses to change visibility (public/private) or default_branch without explicit override. Use when the user says "polish my repos", "update GitHub descriptions", "tag my repos for discoverability", "fix repo metadata", or before any campaign that drives traffic to GitHub. Single-command alternative to clicking through every repo's settings page.
headless-claude
Run the `claude` CLI fast, with your hooks skipped, and unattended for automation — wraps `claude -p` with --setting-sources project (skip your user hooks + heavy auto-memory, ~2s not ~15s), --permission-mode bypassPermissions (no interactive approval prompt that would block a script), and a neutral cwd. Uses your own plan, no API key. Use when the user says "run claude headless", "scriptable claude", "claude -p keeps hanging / prompting", "call claude from a cron/agent", or is automating Claude Code.
live-code-watch
Watch a git repo's working diff get written live — a terminal that redraws ONLY when the code changes (no idle flicker), with brand-new files shown inline via intent-to-add. Point it at a repo while an AI agent (or you) edits and watch the raw code appear line by line. Zero deps beyond git. Use when the user says "watch the code", "show me the live diff", "watch this repo change", or wants to see edits happen in real time.
marketplace-publish
Unified marketplace publisher for VS Code Marketplace (via vsce), iOS App Store / TestFlight (via Expo EAS submit), and Google Play Console (via Expo EAS submit). Single command surface around publish.sh + submit.sh. Pre-flight validates credentials before doing anything destructive. --dry-run by default skips uploads. Use when the user says "publish the extension", "submit to App Store", "ship to Play Store", "marketplace publish", "release the vsix", "build and submit", or before any app-store push.
partnership-email-blast
Send per-partner email campaigns from a YAML roster through Resend, rendered via a clean HTML template. YAML is canonical; each partner gets one templated email with {{var}} substitution. Idempotent — re-runs skip partners already sent (unique index on campaign_id + partner_id WHERE ok=true). --dry-run by default, --apply sends, --prod required for live mode. Use when the user says "blast the partners", "fire the launch email", "email the partner list", "outreach campaign", or before a launch announcement.
screen-describe
Ask your local Claude what's on your screen. Captures the macOS screen, hands the image to your own `claude` CLI (your Pro/Max, no API key — BYOK), prints a short description, and deletes the screenshot immediately. Read-only — it looks, never controls. Use when the user says "what's on my screen", "describe my screen", "what am I looking at", or wants an AI second pair of eyes on their display.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.