parallel-web/parallel-agent-skills
10 skills71.1k installs560 starsGitHub
Install
npx skills add https://github.com/parallel-web/parallel-agent-skillsSkills in this repo
1Parallel Deep Researchparallel-deep-research is an agent skill that orchestrates Parallel’s CLI deep-research processor for solo builders who need exhaustive, web-grounded reports instead of fast search snippets. Install it when you already use Claude Code (or similar) with Bash access and want a repeatable three-step ritual: start the run, poll until completion, and download markdown into a named folder. The skill is deliberately expensive and slow—documentation positions it as 10–100x heavier than parallel-web-search—so it should stay behind an explicit user trigger for comprehensive or thorough investigation. It fits indie validation work such as market sizing, competitor landscapes, and technology comparisons, and supports continuation via --previous-interaction-id when a prior research or enrichment thread exists. Output is file-based research suitable for feeding specs, pitch copy, or launch briefs after you skim citations yourself.9.1kinstalls2Parallel Web Searchparallel-web-search is the default agent skill for everyday web research when you use the Parallel agent stack. It tells your coding agent to call parallel-cli search with a natural-language objective, optional keyword queries, JSON output, and a capped result set so you get current information quickly and cheaply. Solo and indie builders use it during idea validation, competitor checks, library comparisons, launch SEO fact-finding, and ad-hoc debugging questions that need up-to-date sources—not a multi-hour deep crawl. You choose a short hyphenated filename and pass it inline to -o under /tmp; the skill documents a concrete React 19 example so agents do not copy placeholder tokens literally. It requires parallel-cli and internet access and is user-invocable with your query as arguments. Treat it as the first stop before writing specs or code that depends on fresh external facts.8.4kinstalls3Parallel Web Extractparallel-web-extract is a Parallel-maintained agent skill that standardizes URL content pulls through parallel-cli instead of burning main-context tokens on large HTML or PDF bodies. Solo builders assembling research pipelines, RAG prep, or implementation notes benefit because extraction runs in a forked subagent context while the parent keeps a compact JSON artifact under /tmp. The workflow is deliberately concrete: pick a short filename stem from the URL topic, substitute it inline in the -o path (not as a shell variable), and pass optional flags when you need focused excerpts, keyword bias, or full page bodies capped by character limits. It targets JavaScript-heavy sites and PDFs called out in the description. You need parallel-cli installed and outbound network access. This is an agent integration skill, not a marketplace MCP server—install policy should follow your org’s rules for third-party CLIs writing to local disk.8.1kinstalls4Parallel Data EnrichmentParallel Data Enrichment is an integration skill for solo builders who need structured web research appended to spreadsheets without hand-copying from tabs. You point parallel-cli at a CSV or inline list of companies, people, or products and specify fields such as CEO name, recent funding, or contact details; the skill walks through optional column suggestion, processor selection, and long-running enrichment jobs that may take several minutes. It supports multi-turn workflows by passing a previous interaction id so a follow-up run carries context from earlier research. The skill assumes parallel-cli is installed and the agent may run Bash scoped to parallel-cli commands with internet access. It fits indie founders validating TAM lists, prepping investor or partnership outreach, or feeding CRM hygiene before lifecycle campaigns. It is not a generic scraper framework—it is the Parallel CLI enrichment path with suggest-then-run ergonomics.8kinstalls5StatusStatus is a thin agent skill for Parallel’s research CLI: given a run ID, it executes a single status command and returns JSON so your coding agent can tell whether background research finished, failed, or is still running. Solo builders use it after kicking off exploratory research during idea or build work, when they do not want to block the session on completion. The skill assumes Parallel is already configured; if the binary is absent, it instructs setup via the parallel-cli-setup flow rather than guessing install steps. It is intentionally narrow—no result fetching, no new runs—making it a dependable polling primitive in multi-step research workflows. Pair it with the skill that starts research runs so your agent can implement wait-and-continue patterns without manual terminal checks.7.9kinstalls6Resultresult is a thin, user-invocable Parallel agent skill that retrieves completed asynchronous research by run ID and formats the JSON for reading in chat. Solo builders who offload deep web research to Parallel’s CLI use it at the moment a background task finishes—competitor scans, market sizing, or source gathering—without re-running the whole job. The workflow is intentionally minimal: pass the run ID, poll with JSON output, then summarize. Because research underpins both early ideation and later validation checks, the skill is tagged multi-phase with Idea research as the primary shelf. It assumes parallel-cli is installed and authorized; if not, the skill routes the user to the setup skill. Ideal pairing is after starting a research task elsewhere in the Parallel skill pack, when you only need the payload.7.9kinstalls7SetupSetup is a phase-specific agent skill for installing the Parallel plugin CLI and confirming authentication before any Parallel-backed automation runs. Solo builders using Claude Code or similar agents invoke it when they first adopt Parallel or move to a new machine. The skill defers install method choice to official Parallel documentation while walking through login, auth check, PATH fixes, and later updates. That reduces friction where agents otherwise guess package managers or skip verification steps. It is narrowly scoped: no research or ship gates—just reliable local integration. After success, downstream Parallel skills or agent flows can assume `parallel-cli` is on PATH and an authenticated session exists.7.5kinstalls8Parallel MonitorParallel Monitor is an agent skill that wraps the parallel-cli monitor command so solo builders can track recurring web changes—prices, SEC filings, competitor pages—and inspect or delete monitors from the terminal. Monitors persist server-side and emit events when something shifts, with optional webhooks for downstream automation. Use it when users ask to monitor, track changes, watch, or alert on web updates, or when they need to list monitors and review event history. It depends on parallel-cli 0.3.0 or newer and internet access. The skill fits early competitive intelligence in Idea and Validate, and ongoing competitive or market watching in Grow. It is an integration workflow, not a scraper authoring framework; you still define what to watch in natural language and CLI arguments.6.9kinstalls9Parallel Findallparallel-findall is an agent skill for solo builders who need curated entity lists—not blog posts or long-form reports. When you ask to list every roofing company in a city, every Series A AI startup in a year, or YC batch dev tools, the skill drives parallel-cli’s entity discovery so your agent works from rows of entities you can sort, filter, and paste into spreadsheets or CRM workflows. It complements parallel-web-search for pages and parallel-deep-research for narrative analysis. Install it when validation and growth work depend on repeatable list-building: prospecting, competitive landscaping, partnership scouting, or investor mapping. You need parallel-cli 0.6.0 or newer and network access; if commands are missing, the skill instructs update via parallel-cli update or pipx upgrade parallel-web-tools.6.9kinstalls10Parallel Cli SetupParallel CLI Setup is a maintenance-oriented agent skill that walks you through installing or upgrading parallel-cli so research workflows do not stall on environment issues. It starts by checking whether the binary is on PATH, then recommends one of several installers—Homebrew on macOS, uv, npm, or pipx on cross-platform setups—and verifies you are on at least version 0.4.0, with upgrade commands tied to how the binary was originally installed. The skill is explicitly user-invocable and grants tightly scoped shell tooling so Claude Code or similar agents can run the diagnostics for you. For solo builders using Parallel for market scans, competitor pages, or deep research passes during early discovery, this skill is the gate before any search or extract command works. It is not a research methodology itself; it is the integration layer that keeps CLI auth, balance, and skill bundles current.383installs