Perplexity Computer Skills: Turning AI from "One-off Chats" into a "Reusable Workflow Engine"

2026-03-16 | ProductHunt · Official Help Center · Changelog

30-Second Quick Judgment

What is it?: Skills is the "reusable instruction set" feature for Perplexity Computer. You teach it once how to do something (like writing a competitor report, making a PPT, or running data analysis), and it will automatically recognize related tasks and execute them according to your instructions. Simply put, it turns your repetitive prompts into a saved, auto-triggering "Playbook."

Is it worth watching?: Yes, but don't rush in. This feature is an enhancement of the Perplexity Computer platform, which is the core—a $200/month multi-model orchestration system. Skills evolves it from "one-off chats" into a "programmable AI workflow engine," but the high pricing and credit-burning issues mean it's not for the average user yet.

Three Questions That Matter

Does it matter to me?

• Who is the target user?: "Power users" who use AI for repetitive work—content creators (repeating the same research → writing → formatting flow), product managers (regular competitor analysis), developers (running the same code review flows), and enterprise teams (standardized report generation).
• Am I the target?: If you type similar prompts into AI tools more than 3 times a week, you are. If you just ask occasional questions, you aren't.
• When would I use it?:
  • Generating weekly competitor reports → Create a "competitive-analysis" Skill.
  • Repetitive data analysis + visualization → Chart Skill generates charts automatically.
  • Research → Writing → PPT pipeline → Chain multiple Skills together.
  • Import SKILL.md from Claude Code → Migrate workflows across platforms.

Is it useful for me?

ROI Judgment: If your daily work is heavily dependent on AI and you spend $100+/month on AI tools, the "configure once, use forever" nature of Computer Skills can save massive amounts of time. However, if you mainly use Free or Pro versions for simple searches, the $200/month barrier is too high.

Is it exciting?

The "Wow" Factors:

• Auto-recognition and Loading: Say "Make a PPT," and the Slides Skill activates; say "Do a research report," and the Research Skill starts—no manual selection needed.
• Cross-platform Migration: Drag and drop SKILL.md files from Claude Code or Codex to use them immediately without rewriting.
• Multi-Skill Collaboration: One task can automatically call multiple Skills in a chain, such as Research → Report → Slides, fully automated.

Real User Feedback:

"burnt 2,000 credits on Perplexity Computer to build an alpha ai news dashboard... product looks promising but VERY expensive (you get 10,000 credits for $200/month)" — @EXM7777 (408 likes)

"Perplexity Computer is now available for Pro subscribers." — @perplexity_ai (2,688 likes, 464K views) — Official announcement on March 12 that Pro users can now access it.

For Independent Developers

Tech Stack

• Platform Architecture: Cloud-based multi-agent orchestration system.
• AI Models: Access to 19+ models—Claude Opus 4.6 (core reasoning), Gemini (deep research), GPT-5.2 (long context), Grok (light/fast tasks), Nano Banana (images), Veo 3.1 (video).
• Skills Format: Markdown files + YAML frontmatter (name/description fields), fully compatible with Claude Code SKILL.md.
• Execution Environment: Cloud sandbox; every task starts in a clean environment.
• Integrations: 400+ app connectors (Gmail, Slack, Notion, GitHub, Salesforce, etc.).

Core Implementation

Skills uses a three-layer loading architecture: Level 1 is the Trigger Layer—the description field determines when to activate the Skill (this is crucial; official docs say "if a Skill doesn't trigger, 99% of the time it's a bad description"); Level 2 is the Execution Playbook—specific step-by-step instructions; Level 3 is the Reference Library—examples and templates. The system automatically matches and loads the most relevant Skill combination based on user input semantics.

The SKILL.md format has been standardized to v0.9 by the Agentic AI Foundation (supported by the Linux Foundation), and adopted by Microsoft, OpenAI, Anthropic, and Cursor. This means a SKILL.md you write can migrate seamlessly between Perplexity Computer, Claude Code, OpenAI Codex, and VS Code Agent Skills.

Open Source Status

• Is it open?: The platform is closed, but the SKILL.md format is an open standard.
• Similar Open Source Projects: perplexity-super-skills (a community-maintained collection of Skills across 11 domains like Dev, Marketing, Sales, Finance, Legal, PM, Ops, Research, and Content).
• Difficulty to build: Low—if you have a Claude Code SKILL.md file, just import it. Creating one from scratch via dialogue takes only a few minutes.

Business Model

• Monetization: Subscription + Credit consumption.
• Pricing: Max $200/mo (includes 10,000 credits), Pro $20/mo (includes 4,000 bonus credits), extra credits billed based on usage.
• User Base: Perplexity handles 30 million queries daily, with 20%+ monthly growth.

Giant Risk

While this is a Perplexity-specific enhancement, the "Skills" concept isn't exclusive—Claude Code, OpenAI Codex, and Microsoft VS Code Agent Skills have similar mechanisms. Perplexity's differentiator is the "Multi-model Orchestration + Skills" combo, not the format itself. The risk: if Anthropic or OpenAI adds multi-model orchestration to their own agents, Perplexity's moat narrows. However, currently, no competitor can simultaneously dispatch 19 different models.

For Product Managers

Pain Point Analysis

• What problem does it solve?: Users having to re-explain requirements, formatting preferences, and workflows every time they use AI for repetitive tasks.
• How painful is it?: Medium-to-high frequency; it's a must-have for heavy AI users but a nice-to-have for occasional users.

User Persona

• Core User: Knowledge workers using AI daily (content creators, analysts, developers).
• Advanced User: Enterprise teams needing standardized AI workflows (unified research methods, report formats).
• Use Cases: End-to-end Research → Analysis → Output; repetitive content production; code review and architecture analysis.

Feature Breakdown

Competitor Comparison

Key Takeaways

1. SKILL.md Cross-platform Compatibility: Instead of locking users in, they allow users to import workflows from other platforms—a smart growth strategy.
2. Auto-activation Mechanism: No need to manually select skills; semantic matching loads them automatically, lowering the barrier to entry.
3. Three-layer Loading Architecture: The Trigger → Execution → Reference design is a great reference for any team building "reusable AI workflow" products.

For Tech Bloggers

Founder Story

• Founder: Aravind Srinivas, born 1994 in Chennai, India.
• Background: IIT Madras CS dual degree → UC Berkeley PhD → Worked at OpenAI, DeepMind, and Google Brain (a rare trifecta).
• Why build this?: Co-founded Perplexity AI in August 2022 with Denis Yarats, Johnny Ho, and Andy Konwinski. The goal was a search engine that provides cited answers. Moving to Computer + Skills is a strategic leap from "answering questions" to "completing tasks."
• Achievement: India's youngest billionaire; company valuation at $21.21B.

Controversies / Discussion Angles

• Angle 1: "Are credits the new API call trap?" — Users complain about opaque credit consumption, with cases of 20,000 points burned on one task. $200/mo is just the entry price; the real cost is in the credits.
• Angle 2: "Multi-model Orchestration vs. Single-model Specialization" — Perplexity bets that "no single model is all-powerful," so they use 19. But this means their core capability is built on competitors' models—a high-risk dependency.
• Angle 3: "SKILL.md as the Universal Language of AI Agents?" — With Linux Foundation support and adoption by Microsoft/OpenAI/Anthropic, this could be the "Dockerfile moment" for AI.

Hype Data

• PH Rank: 461 votes.
• Twitter Buzz: Official Pro announcement got 2,688 likes and 464K views; CEO personally responds to billing issues.
• Awesome Agents Rating: 7.2/10.

Content Suggestions

• Practical Angle: "I spent $200/month to use Perplexity Computer Skills for a week—here's which workflows actually changed." (Hands-on review).
• Trend Angle: The standardization of SKILL.md—"The one Markdown file to rule all AI Agents."

For Early Adopters

Pricing Analysis

Getting Started Guide

• Time to start: 5 mins (built-in Skills) / 30 mins (custom Skills).
• Learning Curve: Low (dialogue creation) to Medium (writing SKILL.md).
• Steps:
  1. Subscribe to Pro ($20) or Max ($200).
  2. Go to the Computer interface → Skills tab.
  3. Browse built-in Skills or click "Create Skill."
  4. Describe the task you want to automate; the system generates the Skill.
  5. Or upload an existing SKILL.md file (migrated from Claude Code/Codex).
  6. Next time you input a related task, the Skill activates automatically.

Pitfalls and Complaints

1. Credit Black Hole: A code scanning task burned 21,000 points (double the monthly Max quota); you only find out after it's done. Test with small tasks first.
2. No Real-time Preview: Writing frontend code? You have to deploy to Vercel every time to see changes, taking 2-3 minutes per round.
3. Silent Failures in Sandbox: npm install fails without reporting an error, and the agent loops in the dark burning credits.
4. Unannounced Model Downgrades: During peak hours, you might be switched to a weaker model without notice; the model tag might be misleading.
5. Billing Instability: CEO Aravind Srinivas apologized on March 10, stating "billing eng issues will be fixed today."

Security and Privacy

• Data Storage: Cloud-based with persistent memory (retains preferences across sessions).
• Privacy Risk: Uploaded images can be accessed externally via URL (even without login); security is questionable.
• Audit: No public third-party security audit reports available.

Alternatives

For Investors

Market Analysis

• Sector Size: The AI Agent/Workflow automation space spans RPA ($4.68B → $35.84B by 2033), Low-code ($24.83B+), and AI Search/Assistants.
• Growth Rate: Perplexity's ARR grew from $35M in 2024 to $200M in 2025, with a 2026 target of $656M (approx. 19x growth in two years).
• Drivers: Efficiency demands from knowledge workers, exponential growth in AI capabilities, and fragmented enterprise SaaS workflows.

Competitive Landscape

Timing Analysis

• Why now?: In 2026, AI models have crossed the threshold of "reliably executing multi-step workflows"; SKILL.md v0.9 is being adopted as a standard; the shift from "Chat AI" to "Agentic AI" is happening.
• Tech Maturity: Core models are strong, but the orchestration layer still has issues (opaque credits, poor sandbox debugging).
• Market Readiness: Enterprise demand is clear (standardized workflows), while consumers remain price-sensitive.

Team & Funding

• Founders: Aravind Srinivas (CEO), ex-OpenAI/DeepMind/Google Brain, UC Berkeley PhD.
• Core Team: Denis Yarats, Johnny Ho, Andy Konwinski (top-tier AI/Backend backgrounds).
• Funding: $1.22B - $1.5B+ across 10 rounds.
• Investors: Jeff Bezos, Yann LeCun, SoftBank Vision Fund, NEA, NVIDIA, Samsung, Cristiano Ronaldo.
• Valuation: $21.21B (2026 Series E-6).
• Revenue Multiple: ~106x based on $200M ARR; ~32x based on $656M target.

Conclusion

Final Verdict: Perplexity Computer Skills has pushed SKILL.md from a "niche developer tool" toward a "universal AI workflow standard." The product strategy is spot on—reusable, auto-triggering, and cross-platform—but the $200/month barrier and credit black hole make it suitable only for heavy professional users for now. As Pro users gain access, the barrier is beginning to lower.

Resource Links

2026-03-16 | Trend-Tracker v7.3