vb-tyagi
UserAgent-native X poster — one MCP server for any coding agent (Claude Code / Cursor / Codex / Windsurf). Connect X once, then create/schedule/post from your agent session. Token + casserole guardrail + send are one server-side unit; your agent holds only a short-TTL handle, never the token. "The AI writes; casserole decides what ships."
Categories
Indexed Skills (30)
best-time
Suggest good X post times — a v1 heuristic (audience timezone + known-good windows), honest it's not your data yet. Outputs slots you feed straight to create_loop.
build-in-public
Turn recent git activity into build-in-public X posts, queued to a schedule. You (the agent) draft from the real commits — capx never writes; it ships them guarded.
cadence-planner
Turn a backlog of drafts or a content goal into a non-robotic posting cadence — how many a week, which days and times — then create the loop(s). You draft; capx only ships.
changelog-thread
Turn a commit range or version bump into a 'what shipped in vX' X thread — you (the agent) read the range and write each post; capx never writes, it ships them guarded.
fix-note
Turn a bug you just fixed — the session's debugging or a closed issue — into one specific 'we fixed X' post. You draft from the real diff; capx never writes, it ships it guarded.
gap-alert
A loop is running low — refill it from your newest commits. list_loops finds the near-empty one; you draft fresh grounded posts; capx ships them guarded via top_up_loop.
hook-rewrite
Rewrite the FIRST line of a draft for stopping power without tipping into bait. You draft 2-3 openers grounded in the real content; casserole still gates at send.
image-director
Direct the user's OWN image-gen tool to a high-quality, on-brand image for a post: pick the model, engineer the prompt, generate, then upload + attach. capx runs no model — it orchestrates yours and ships the post guarded (caption only).
launch-thread
Turn a README + release into a launch thread: hook, problem, solution, proof, CTA. You draft every part from the repo's reality; capx never writes — it ships each guarded. Heavy approval before live.
media-connect
Guide the user to connect their OWN media-generation MCP servers (higgsfield, fal, kling, ...) to their agent, so capx can attach the images/videos those tools produce. capx generates nothing — it orchestrates your tools and ships the result.
model-guide
Reference: which image/video model to reach for by job (photoreal, illustration, product shot, short clip) and the tradeoffs, so you choose well BEFORE generating. capx integrates no provider — you generate with your own connected tools.
quickstart
First-time walkthrough — check the connection, link X if needed, then help the user post their first thing and read the guardrail verdict. Warm, one step at a time.
reformat
Reshape a draft you already wrote — long post to a thread, a thread to one tight post, or fit it to X's norms. You reshape; the meaning stays; capx ships it guarded.
repurpose
Turn a blog post, README, or long doc (a file path or URL) into an X thread. You (the agent) read it and write the posts; capx never generates — it ships them guarded. Post now or queue.
self-host-guide
Stand up your OWN capx chokepoint — the self-hostable image that holds the token vault and runs the guard at send. Walks the real needs from the repo; never fabricates commands.
ship-note
Turn a merged PR or GitHub release into a clear "we shipped X, here's what it means for you" X post or short thread. You draft from the real diff/notes; capx ships it guarded.
thread-builder
Turn raw material into a numbered X thread — a sharp hook, one idea per post, a clean close/CTA. You draft it; capx ships each post in order, guarded.
til
Turn the one non-obvious thing you learned this session into a single sharp TIL post. You (the agent) write it from real session context; capx never writes, it ships it guarded.
video-director
Direct the user's OWN video-gen tool to a short, X-native clip — hook in the first second, right aspect ratio, length norms — then upload + attach. capx generates nothing; it orchestrates your tool and ships the post guarded (caption only).
voice-match
Learn the developer's own X voice from their recent posts — tone, length, emoji/hashtag habits, rhythm — so other capx skills draft in it. A read-only look at their own timeline; capx never writes.
connection-health
Check the capx connection and self-heal it — whoami reports your @handle, lane, and needs-reauth; if it's stale or nothing's connected, it routes you to connect.
draft-review
Dry-run a draft through the casserole guardrail WITHOUT posting — see whether it would pass, be held, or be blocked, and why, so you can fix it before it ships. You draft; capx just lints.
prompt-engine
Turn a vague creative ask into a precise, model-optimized prompt for image or video generation — the quality engine the director skills call. It produces a PROMPT; capx never generates the media itself.
super-extraction
Forensic website extraction — tear down any website, web app, or landing page into a full local source mirror + reference screenshots + a TEARDOWN.md complete enough that an engineer who has never seen the site could rebuild it pixel-for-pixel. Curl-first capture, deep code read (every constant, shader, easing, breakpoint), live browser verification, licensing flags, 1:1 rebuild checklist. Triggers: "extract this website", "tear down / teardown this site", "reverse engineer this landing page / app / UI", "break it down pixel by pixel", "pick it apart piece by piece", "replicate this design/site" (extraction is always step 1 of replication), "super extraction", "every inch and pixel", "how does this site work under the hood". NOT for: simple screenshots or scraping text (that's vbt-browse-op); building UI from a taste spec (a UI-builder skill); aesthetic-only decoding (a design-taste skill) — this is engineering forensics: code + behavior + assets, not just visual taste. Deliverable = TEARDOWN.md + full local
browse-op
Headless browser automation. Use when the user asks to open a URL, check a website, scrape data, fill a form, test a page, read web content, take a screenshot, interact with a web app, or any task requiring browser access. TRIGGER when: user mentions a URL, asks to "open", "browse", "check", "visit", "navigate to", "scrape", "read the page", "screenshot", "test in browser", "fill the form", "click", "what does [site] look like", or any web interaction. DO NOT TRIGGER when: user explicitly asks to use Claude in Chrome (visible browser), or the task is purely about reading local files.
superprompt-debloat
Prompt assessor and surgeon for EXISTING prompts. Takes a prompt the user already has (pasted text, a file path, or several prompt files), audits it line-by-line for bloat, weak wording, ambiguity, contradictions, and missing load-bearing pieces, then delivers a tightened, strengthened rewrite plus a change log explaining every cut, edit, and addition. Never executes the prompt — the improved prompt IS the deliverable. Use when the user invokes /superprompt-debloat, or asks to "debloat this prompt", "assess/audit my prompt", "tighten this system prompt", "improve/edit this existing prompt", "is this prompt bloated". Twin of superprompt — that skill engineers a fresh task prompt and RUNS it; this one operates on already-written prompts and hands them back better. Do not trigger when the user wants a prompt executed.
superprompt
Active prompt-engineering layer. Takes the user's raw prompt, decodes the real intent, rewrites it into a heavily-engineered enhanced prompt, shows that enhanced prompt, then IMMEDIATELY starts executing it in the same session — no confirmation step, as if the user had typed the enhanced version themselves. Use when the user invokes /superprompt (with their raw prompt as arguments), or explicitly asks to "enhance this prompt and run it", "supercharge my prompt", "prompt-engineer this then do it", "make my prompt better and go". Works for any task type — research, coding, writing, analysis, planning, brainstorming. Do NOT self-trigger on ordinary requests where the user hasn't asked for prompt enhancement — this is an on-demand amplifier, not an always-on filter.
vbt-font-emulator
Emulate a typeface from a reference screenshot and generate production SVG wordmarks/typography for websites, slides, and posters. Use when the user shares a font/type/lettering/typography reference image and wants text written in that style, or asks to "emulate this font", "write X in this type", "make a wordmark like this", "recreate this lettering as SVG", "custom type from this screenshot", or wants new words added to a previously emulated style. Draws ORIGINAL vector letterforms capturing the reference's DNA (never traces, never pirates the font), assembles words via the bundled engine + glyph libraries, and verifies every render visually before delivery. NOT for choosing/installing existing fonts, full UI builds (vbt-ui-builder-from-dtp), or general design-taste decoding without a type deliverable (vbt-design-taste-profile).
self-code-review
Critical self-review of code you just wrote or are about to ship. Use after completing a coding task, before committing, or when the user says "review your code", "check your work", "audit this", or "is this clean?". Produces an honest assessment with a concrete improvement plan — only if improvements are warranted. Never forces changes for the sake of appearing thorough.
next-1t-thesis
Pressure-test or shape AI startup ideas against the Sequoia "Services - The New Software" thesis. Use when the user shares an AI business idea (fuzzy or formed), asks whether something is a copilot or an outcome business, wants to validate/stress-test/transform an AI thesis, mentions services-as-software, or references Tier 1/Tier 2/$1T/Sequoia services thesis. Three modes - Discovery (shape a fuzzy idea), Litmus (stress-test a formed idea), Transformation (mutate a weak idea into stronger forks). Refuses to flatter, forces concreteness, always names the Tier.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.