10 Claude Skills I Can’t Live Without

Summary: Reusable “Claude Skills” (MD-file automation)

The video argues that in 2026, the main advantage of Claude Skills (instead of basic chat) is reusable, instruction-based automation. Rather than repeatedly providing the same context, you create/install a skill as an MD file that Claude can run again and again via a command.

---

What “Claude Skills” are (core concept)

• Claude Skills are MD files containing instructions Claude can follow repeatedly.
• You run a skill in Claude by typing: / + the skill name.
• The creator suggests this can replace “building agents” for many workflows.

---

10 Claude skills shown (key features and use-cases)

1. Skill Creator (official Anthropic skill)

   • Used to generate new Claude skills interactively.
   • Asks questions and then produces the skill’s MD file.
   • Example: “YouTube brand scout”
     • Scans one or more YouTube channels (via Chrome) to identify sponsored videos from the last ~6 months.

2. HookForge

   • Generates strong content hooks from pasted content (e.g., to improve videos/posts).
   • Uses psychological angles such as:
     • Curiosity (e.g., “quietly replaces $200/month worth of subscriptions”)
     • Contrast (reframing “saving money” as a “cheaper plan” misconception)
     • Controversy (“MidJourney in 2026 is a skill issue…”)
     • Social proof (“top 1%… secret stack… tools free on GitHub”)
   • Claimed benefit: accelerates ideation work that would otherwise require a team.

3. Remotion skill (launch/trailer video generation)

   • Uses Remotion to create a trailer-like video in minutes (video claims ~8 minutes).
   • Example: generates a brand video using assets from the user’s website (e.g., “Hello isan.com”).

4. Humanizer

   • Rewrites AI-generated drafts to sound more human and less “generic AI slop.”
   • Produces templates (example: sponsorship outreach email).
   • Workflow support is included—after generating, the presenter mentions sending via Gmail.
   • Output is described as shorter, direct, and “warm/confident.”

5. Whisper Flow (dictation-to-prompt structuring)

   • Lets you speak prompts and converts them into a cleaner, ready-to-use instruction format.
   • Claimed advantage vs native dictation:
     • better formatting
     • fewer messy/duplicative outputs
     • more structured prompts
   • Example result: a concise email request for a podcast.

6. Infographic Builder

   • Creates interactive infographics for LinkedIn/Instagram.
   • Example described: an infographic explaining psychological tactics brands use to drive purchases, with:
     • hover/interactivity (“moves”)
     • subtext
     • scrollable detailed sections

7. Karpathy Guidelines helper (for Claude Code projects)

   • Applies Andrej Karpathy-style coding guidelines to reduce errors and inefficiency.
   • The guideline components listed include:
     • Think before coding
     • Don’t assume / don’t hide confusion
     • Surface trade-offs
     • Simplicity first
     • Surgical changes
     • Touch only what you must
     • Clean up only your own mess
     • Goal-driven execution (define success; loop until verified)
   • Claimed benefit: improves code quality and saves credits/tokens by avoiding repeated context.

8. Caveman

   • A “credit/token saver” that makes Claude respond with short, essential summaries.
   • Example: summarizing Psychology of Persuasion into key lessons instead of a long response.
   • Claimed savings: up to 75% tokens/credits.

9. Expand & Contract

   • Helps brainstorm and refine a business idea by expanding into a plan and iterating.
   • Example business idea: hiring/“house help” for wealthy people in India.
   • The skill asks structured questions (services, must-haves, GTM/operational must-haves, positioning), then outputs:
     • services scope
     • watch-outs
     • a visual business outline
   • Claimed outcome: deep business clarity in a few minutes.

10. FindK Skills

    • A meta-skill that searches/installs/suggests relevant skills for a given task.
    • Uses skills.sh to match a user’s problem statement to appropriate skills.
    • Example: writing a LinkedIn post about news/trends—selects skills to generate hooks and the post.

---

Additional decision aid highlighted

• Decision Framer (called out as a skill for better decision-making)
  • Takes a decision question and creates structured pro/cons and criteria-based reasoning.
  • Claims to outperform “vanilla” ChatGPT responses by:
    • thinking step-by-step
    • providing a recommendation (with options like “primary with backup,” “keep all,” etc.).

---

Reviews / practical guidance emphasized

• Best practices suggested:
  • Use skills to avoid re-explaining context.
  • Install and reuse prebuilt (or generated) skills via commands.
  • Use Humanizer and dictation structuring to improve output quality (tone + formatting).
  • Use Caveman and Karpathy Guidelines to manage credits/tokens and improve efficiency.
  • Use FindK Skills to discover the right workflow skill for a task.

---

Main speakers / sources

• Speaker: Isan Sharma (host/presenter)
• Company/product sources mentioned:
  • Anthropic (for Skill Creator and the Claude skills ecosystem)
  • Andre Karpathy (Karpathy guidelines referenced)
  • Remotion (video generation skill)
  • skills.sh (skill discovery platform)