prashant-cr
UserAgent Skills for careers, scraping, markets, documents and health — make a resume survive ATS parsing, audit a site before you scrape it, parse any PDF into a spreadsheet, deep-dive a stock, rate an IPO or crypto on live data, plan a diet. Works with Claude Code, Cursor, Copilot and 20+ agents. Install: npx skills add prashant-cr/skills
Categories
Indexed Skills (11)
fix-broken-scraper
Diagnoses why a working web scraper or crawler broke and repairs it one verified step at a time — separates stale CSS selectors and site redesigns from real blocking, 403/429 responses, expired sessions, changed URL routing, and content that moved into embedded JSON or client-side rendering. Use this whenever a scraper, spider, crawler or parser that used to work now returns empty results, None, an empty list, zero rows, HTTP 403/429, a CAPTCHA page or wrong data — whenever someone says their scraping script stopped working, broke overnight, worked yesterday, is suddenly blocked, or their BeautifulSoup, Scrapy, Playwright, Puppeteer or requests code returns nothing.
ipo-rating
Rates an upcoming or open IPO and says whether to apply — scoring the business, valuation against listed peers, issue structure, promoters and disclosure quality, then giving two separate verdicts — one for listing-day gain and one for holding the stock. Reads pre-listing demand signals including grey market premium (GMP) where it exists, subscription figures and the anchor book, and maps the post-listing lock-in expiries. Works on any market. Use whenever the user asks whether to apply for, subscribe to or skip an IPO, wants an IPO analysis, rating or review, mentions GMP, grey market premium, listing gains, allotment, price band, RHP, DRHP or a new listing, asks how much an IPO might list up or down, or names a company that is about to go public — including bare asks like "is this IPO worth it", "should i apply", or "what's the GMP looking like".
news-stock-impact
Turns today's news into a ranked shortlist of listed companies most exposed to it — naming the event, the transmission mechanism, how much the stock has already moved on it, and what would invalidate the idea. Works on any market (US, UK, India, Japan and others) and any market cap. Use whenever the user asks which stocks or sectors will move on the news, what to buy, watch or trade tomorrow, who benefits from a policy change, rate decision, oil move, earnings surprise or geopolitical event, or mentions today's headlines alongside stocks, shares, tickers, equities, markets or investing — including bare asks like "what's moving tomorrow", "any stock ideas from today's news", or a pasted headline with "how does this affect the market".
pdf-parsing
Parses any PDF into structured usable data — classifies the document first (text layer, scanned, fillable form, encrypted), extracts text, tables and form fields with whatever toolchain is actually installed, OCRs scans that have no text layer, and writes a real multi-sheet Excel workbook, CSV, JSON or document without needing pandas or openpyxl. Handles one file or a whole folder into a single spreadsheet with a source-file column plus a list of what failed and why. Use whenever the user wants to parse, read or extract a PDF, pull tables, invoices, bank statements, bills, payslips, receipts, forms or report figures out of one, convert a PDF to Excel, xlsx, CSV, Word or JSON, asks why a PDF returns empty text or scrambled columns, mentions a scanned PDF or OCR, or has a folder of PDFs to turn into a spreadsheet.
scrape-feasibility-audit
Audits a public website to determine how hard it will be to scrape and what to build it with — identifies bot detection (Cloudflare, Akamai, DataDome, HUMAN/PerimeterX, Kasada, Imperva, AWS WAF), CAPTCHA types (reCAPTCHA, hCaptcha, Turnstile, Arkose, GeeTest), robots.txt rules, rate limits, and whether content is server- or client-rendered, then recommends open-source libraries. Use this whenever the user asks whether a site can be scraped, why their scraper is getting blocked or returning empty pages, what anti-bot protection a site uses, which scraping library or framework to pick, or is planning any data-collection or crawling project — even if they only mention a URL and extracting data from it.
stock-deep-dive
Full fundamental analysis of a single listed company, the way a fund manager builds a position — business model, moat, financial and earnings quality, management and capital allocation, what the current price already implies, and a scored 0-10 verdict across six dimensions with bull/base/bear scenarios. Use whenever the user names a stock or ticker and asks whether to invest, buy, hold or avoid, wants a deep dive, fundamental analysis, valuation, research report or second opinion on a company, asks if something is overvalued, undervalued, a good long-term bet or what the future holds for it, is comparing two companies, or is reviewing a position they already own — including bare asks like "thoughts on NVDA?", "is Reliance worth buying", or "analyse this stock for me".
structured-data-extraction
Extracts data from web pages without writing CSS selectors or XPath, by reading the machine-readable structures a page already publishes — JSON-LD, schema.org microdata, Open Graph tags, embedded __NEXT_DATA__ and application/json state blobs, HTML tables and RSS feeds — and locating a value by name so you get a JSON path instead of guessing at DOM structure. Use this whenever someone is writing selectors to scrape a page, asks how to get product prices, article metadata, listings, ratings or tables off a site, wants extraction that survives redesigns, is tired of selectors breaking, or is about to reach for BeautifulSoup, parsel or a headless browser to pull fields out of HTML.
weight-loss-diet-plan
Builds a personalised fat-loss diet plan — a full template day and a full seven-day menu, sized to the person's own calorie and protein numbers after a short intake on lifestyle, current eating pattern, cooking time and food preference (vegetarian, vegan, eggetarian, non-vegetarian, pescatarian, halal, jain and others), with a shopping list and the rules for what to change when the scale stalls. Use whenever the user asks for a diet plan, meal plan, weight loss or fat loss plan, weekly menu, how many calories they should eat, how much protein they need, what to eat to lose weight, wants their current diet reviewed or fixed, or says their weight loss has plateaued — including bare asks like "plan my diet", "i want to lose 10 kg before december", "what should i eat tomorrow", "is my meal plan enough protein" or "my weight is stuck".
crypto-rating
Rates a crypto asset on live data only and says whether to buy it and at what size — never from remembered prices, because in crypto a stale figure is wrong by a multiple rather than a little. Fetches and cross-checks the current price, market cap, supply and drawdown, screens new or thin tokens for rug and trap patterns before rating anything, does the dilution and unlock arithmetic, scores six dimensions, then turns the user's own capacity for loss into a maximum position size with the crash scenarios shown in money. Gives two separate verdicts, long-term hold and near-term entry, because they frequently disagree. Use whenever the user names a coin or token and asks whether to buy, sell, hold or avoid it, wants a crypto rating, analysis, second opinion or price view, asks how much of something to buy or how much to put in, mentions tokenomics, unlocks, FDV, altcoins, memecoins or whether a token is a scam, or asks about sizing and risk in their crypto portfolio.
proxy-finder
Works out what proxy or scraping infrastructure a specific site actually needs and what it will really cost per month — probes the target for its bot-defence posture and page weight, checks robots.txt for the paths you want, then prices every option including the ones that are not proxies at all (official API, scraping API, data vendor) on total cost of ownership with retries and engineering time included. Fetches vendor pricing live and refuses to rank a price it could not verify, because proxy rates change constantly. Use whenever the user asks which proxy or proxy provider to use for a site, wants the cheapest or most reliable proxies, asks about residential versus datacenter versus mobile IPs, rotating IPs, proxy pricing per GB, scraping APIs or unblockers, says their scraper is getting blocked and asks what to buy, wants to compare proxy vendors or cut proxy costs, or asks what infrastructure a scraping project will need and what the monthly bill looks like.
value-for-money
Finds the product that is genuinely the best buy for one person's use — asks what the thing is actually for, searches live listings, then ranks candidates on cost of ownership rather than sticker price, with ratings confidence-adjusted for how many reviews back them and manufactured-looking review distributions excluded rather than rewarded. Quotes no price it did not fetch, and is willing to answer that the right move is last year's model, a refurbished unit, waiting for a known sale, or not buying. Knows Indian retail specifics like seller warranty, GST invoice, grey imports and festive cycles, and asks the market first. Use whenever the user wants help choosing what to buy, asks which product is best or best value, wants the cheapest good option, a recommendation under a budget, a comparison between two or three products, whether something is worth the money or worth upgrading to, whether a deal or discount is real, or whether a product's ratings and reviews can be trusted.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.