メニュー
Use for web scraping, crawling, document extraction, API parsing, or building validation-heavy data pipelines using Firecrawl or local Python scripts.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Default entry point for any research request — a hybrid router that classifies the question deterministically and either delegates to a specialist research skill (pulse for trends/sentiment, grants for NIH funding, litreview for academic literature, syllabus for course reading, patent for prior-art + IP landscape, dossier for entity research) or runs its own plan-decompose-multi-source-search-synthesize-cite fallback workflow when no specialist matches. Always surfaces the routing decision so users can override. Use when the user makes any research request that doesn't obviously match a more-specific specialist skill (e.g., "research [topic]", "look into [topic]", "what do we know about [topic]", "investigate [topic]", "find me information on [topic]", "do some research on [topic]", "I need to understand [topic]"). Output is a markdown briefing (default) or .docx document (on request) with full citations and an audit log.
Run a disciplined, multi-source research investigation for a high-stakes question or decision — fan-out web search across many channels, parallel sub-agents, source triangulation (each claim backed by ≥3 independent sources), an adversarial review pass, and every source saved to its own file with verbatim quotes for reuse. Use when a low-quality answer is expensive: strategy work, comparing N products/methods/markets, validating a hypothesis with external data, or mapping how a field works. NOT for quick fact-checks (answer directly), structured 12-dimension competitor scoring (use competitive-teardown), or fast topic overviews where the decision risk is low (use the research router instead).
Manage local SEO for service-area businesses — appliance repair, HVAC, plumbing, cleaning, and any business that serves customers at their location. Use when the user wants to: audit Google Business Profile, generate neighborhood service area pages, check NAP consistency across directories, create LocalBusiness schema, or write review responses. Triggers: 'local SEO', 'Google Business Profile', 'GBP', 'service area page', 'NAP consistency', 'local citations', 'LocalBusiness schema', 'review responses', 'Google Maps ranking'. NOT for national SEO (use seo-audit). NOT for general schema (use schema-markup). NOT for AI answer-engine visibility (use aeo).
Central router for the marketing skill ecosystem. Use when unsure which marketing skill to use, when orchestrating a multi-skill campaign, or when coordinating across content, SEO, CRO, channels, and analytics. Also use when the user mentions 'marketing help,' 'campaign plan,' 'what should I do next,' 'marketing priorities,' or 'coordinate marketing.'
Runs a disciplined Discuss -> Map -> Decompose -> Execute -> Verify loop that grounds code in verified structure — no invented APIs, no assumed imports, no placeholder code — with a lazy-senior-dev YAGNI ladder that deletes unnecessary code before it is written. Use when a coding task is high-stakes, complex, or spans existing code (auth, databases, migrations, multi-file features), or when the user explicitly asks to plan carefully before coding, avoid hallucinated code, or work rigorously. Not for trivial edits, typos, or throwaway one-off scripts — those do not need the full loop.
Mint a tamper-evident, post-quantum-signed receipt for a consequential agent action (deploy, delete, pay, grant-access, model decision) so it can be verified later from the certificate alone. Use when an autonomous agent takes a side-effecting action that may need to be proven later, or when satisfying EU AI Act Article 12 record-keeping. Three decisions: whether an action needs a receipt, minting it, verifying it. Signing is delegated to the open-source OpenAgentOntology package. Not after-the-fact log analysis; not a hosted notary; not a legal opinion.
| name | universal-scraping-architect |
| description | Use for web scraping, crawling, document extraction, API parsing, or building validation-heavy data pipelines using Firecrawl or local Python scripts. |
Design complete, robust data-extraction pipelines with intelligent routing, validation, and token-budget tracking — not brittle one-off scripts.
Dependency Notice: BYOK (Bring Your Own Key) pattern for Firecrawl; API keys must only be loaded via environment variables. Per-script dependencies:
| Script | Dependencies | Exact CLI |
|---|---|---|
scripts/validate_extraction.py | stdlib only | python3 scripts/validate_extraction.py output.json --json |
scripts/firecrawl_example.py | firecrawl, requests (template; --sample runs offline) | python3 scripts/firecrawl_example.py --sample |
scripts/local_bs4_example.py | beautifulsoup4, pandas (template; --sample runs offline) | python3 scripts/local_bs4_example.py --sample |
Check for context first:
If project-context.md exists, read it before asking questions. Determine the target data format, scale of extraction, and deployment environment before writing any code.
This skill supports 3 extraction modes based on intelligent routing:
Use when the source is a public URL, heavily dynamic (JS/SPA), requires search-first discovery, or involves bulk crawling across a domain.
Use when extracting from local files (PDF, Excel, CSV), the data is private/sensitive, or the target is a simple static HTML page where Firecrawl is overkill.
Use when Firecrawl handles URL discovery/web extraction, but local Python (Pandas) is required to clean, normalize, and structure the output before saving.
When executing a scraping task, always follow this sequence:
scripts/firecrawl_example.py (Mode 1) or scripts/local_bs4_example.py (Mode 2); run each with --sample first to see the expected summary shape without network access.python3 scripts/validate_extraction.py extracted_output.json --json on every extraction result before delivering it. It exits 0 only on {"status": "ok"}; warning (empty output) or error (malformed JSON) exit 1 — fix and re-extract, never ship unvalidated data. Beyond this structural gate, also check required fields and duplicates against the pipeline spec before delivering.Surface these issues WITHOUT being asked when you notice them in context:
os.getenv('FIRECRAWL_API_KEY').| When you ask for... | You get... |
|---|---|
| "Scrape this site" | A fully validated Python extraction script with routing logic and error handling. |
| "Get data from this table" | A clean CSV/JSON dataset with a summary log of row counts and empty values. |
| "Crawl these docs" | A Markdown deliverable chunked for LLM token limits. |
div > span > ul > li:nth-child(3)). Use data attributes or robust structural anchors.robots.txt or implementing sensible rate limits.