Agent Skills

repository·main·Indexed 41 days ago

https://github.com/samber/cc-skills

A repository of reusable, domain-specific instruction sets designed to extend AI assistants with expertise in Engineering, Marketing, Developer Relations, and Productivity. Compatible with tools such as Claude Code, Openclaw, Gemini CLI, Cursor, Copilot, OpenCode, Codex, and Antigravity. Features atomic skill units with progressive disclosure to prevent context bloat and includes performance metrics based on adversarial testing.

Tokens
300.1K
Snippets
371
Records
1K
Agent score
52%

What's inside Agent Skills

  1. What the Weekly SEO agent does

    main

    The agent performs 12 automated tasks in a single weekly pass to generate a comprehensive SEO report:

    1. Ranking changes: Compares GSC and Ahrefs data to flag significant drops in position, impressions, or CTR.
    2. Page-2 opportunities: Identifies queries in positions 11-20 and suggests content edits based on top-ranking competitors.
    3. New backlinks: Pulls new referring domains from Ahrefs and categorizes them (High Authority, Niche, Neutral, or Spam).
    4. Lost backlinks: Identifies lost domains and drafts outreach emails for high-value (DR > 40) losses.
    5. Core Web Vitals drift: Monitors GSC and PageSpeed Insights for regressions in LCP, INP, or CLS.
    6. Indexation health: Summarizes GSC indexing reports by reason (e.g., 'Crawled - currently not indexed').
    7. AI bot traffic: Uses PostHog to monitor traffic from bots like GPTBot or ClaudeBot, checking robots.txt compliance for sensitive site types.
    8. Competitor SERP monitoring: Tracks the top 3 results for target keywords and flags new entrants or dropped competitors.
    9. Content freshness audit: Identifies aging content with declining clicks and recommends refresh priorities.
    10. Schema validity: Samples pages and validates JSON-LD against schema.org requirements.
    11. Stats memory snapshot: Appends longitudinal data to weekly-seo/memory/stats.csv and weekly-seo/memory/keywords.csv.
    12. Change log validation: Reviews weekly-seo/memory/changelog.md to validate if previous site changes had the intended SEO impact.
  2. Use the copywriting-prose-creator skill

    main

    The copywriting-prose-creator skill is used to codify a person or brand's writing style (lexicon, syntax, rhythm, structure, and signature moves) into a PROSE.md guide. This guide is independent of emotional tone and serves as the system-prompt substrate for downstream writing skills.

    When to use this skill

    • To create a PROSE.md guide from scratch using SOUL.md and TONE.md.
    • To port an existing PROSE.md to a new channel grouping.
    • To reverse-engineer prose patterns from a content corpus.

    When NOT to use this skill

    • Do NOT use this to write actual content. Instead, use:
      • samber/cc-skills@linkedin-ghostwriting
      • samber/cc-skills@technical-article-writer
    • Do NOT use this for tone decisions. Instead, use samber/cc-skills@copywriting-tone-of-voice-creator.
    • Do NOT use this for hooks or CTAs.
    name: copywriting-prose-creator
    description: "Codifies how a person or brand writes — lexicon, syntax, rhythm, structure, signature moves — into a PROSE.md guide, independent of emotional tone."
  3. Use the Copywriting Hooks catalog for content creation

    main

    The copywriting-hooks skill provides a catalog of 18 distinct psychological frameworks designed to capture reader attention at the start of an article, essay, or post. Each hook is categorized by its mechanism, recommended use cases, and pitfalls to avoid.

    Hook Categories Overview

    1. Curiosity Gap: Open an information gap the reader wants to close.
    2. Contrarian: Challenge a consensus belief.
    3. Bold Claim / Promise: State the outcome upfront before providing proof.
    4. Scene Opener / In Medias Res: Drop the reader into a specific moment without setup.
    5. Surprising Statistic: Lead with a number that violates prior expectations.
    6. Question: Pose a question the reader is already implicitly asking.
    7. Quote + Disagreement: Use a known quote as a springboard to twist the narrative.
    8. Personal Confession: Admit something vulnerable to build rapport and universalize the lesson.
    9. Concrete Specific Detail: Replace abstractions with a single, vivid, sensory detail.
    10. Pattern Interrupt: Break the expected reading rhythm with a fragment or sudden shift.
    11. Direct Problem (PAS): Name a specific pain point, sharpen it, and hint at a solution.
    12. Promise / Benefit: State a specific, time-bounded outcome.
    13. Historical Analogy: Use a historical vignette to pivot to a modern context.
    14. Definition Reversal: Reframe a concept (e.g., "X is not Y, it is Z").
    15. Authority Borrow: Lead with a recognizable name performing a specific action.
    16. Time Anchor: Lead with a specific date, hour, or moment to ground the reader.
    17. Conditional: Use "If you... then..." to self-segment the target audience.
    18. Open Loop: Start a narrative thread and withhold the resolution to build tension.
  4. Use the press-release-writer skill

    main

    The press-release-writer skill is designed to draft or improve professional press releases, media announcements, or news releases (e.g., product launches, funding rounds, partnerships, crisis communications, or M&A).

    When to use:

    • When you need to announce something to the press.
    • For product launches, executive hires, or open source milestones.
    • For region-specific announcements requiring specific style guides or datelines.

    When NOT to use:

    • For blog posts or technical articles (use samber/cc-skills@technical-article-writer instead).
    • For newsletter issues (use samber/cc-skills@substack-ghostwriting instead).

    Capabilities:

    • Supports various media targets: print, digital/wire, broadcast, social/SMPR, and trade press.
    • Handles global regional conventions and language requirements (e.g., English, French).
    • Uses an 'inverted pyramid' structure to ensure news impact.
  5. Use the linkedin-ghostwriting skill for B2B content

    main

    The linkedin-ghostwriting skill is designed for generating conversion-focused B2B LinkedIn posts that prioritize results and authority. It is intended for users who want to write LinkedIn content, create ghostwritten posts for founders/executives, or develop a B2B social strategy.

    Note: Do NOT use this skill for newsletter or Substack content; use samber/cc-skills@substack-ghostwriting instead.

    Key Capabilities:

    • Extracting quantified stories and insights.
    • Engineering high-tension hooks.
    • Structuring posts using copywriting frameworks (ABT logic).
    • Formatting content for mobile-first consumption.

    Allowed Tools: Read, Edit, Write, Glob, Grep, Agent, AskUserQuestion

    name: linkedin-ghostwriting
    description: "B2B LinkedIn ghostwriting — hooks, post structures, and copywriting frameworks for conversion-focused posts..."
    user-invocable: true
    license: MIT
    compatibility: Designed for Claude or similar AI agents.
  6. Use the frontend-design-deslop skill to design non-generic UIs

    main

    The frontend-design-deslop skill is designed to prevent the 'AI-slop' look (generic Tailwind defaults, Inter font, soft shadows, indigo/violet color palettes) by enforcing a strategy-first design process.

    When to use it:

    • When building or styling any web frontend (landing pages, SaaS, dashboards, ecommerce, etc.).
    • When the user requests to "de-slopify", "deslop", "make it not look like AI", or "give it character".
    • Even on a bare "build a UI for X" prompt where no aesthetic is specified.

    Core Workflow:

    1. Phase 0 (Discovery): Identify the artifact type, audience, and commit to 3-5 brand adjectives and an aesthetic essence. Always check for an existing DESIGN.md before starting.
    2. Phase 1 (Design System): Translate strategy into typography, OKLCH color systems, and a token table (spacing, radius, shadows, etc.).
    3. Phase 2 (Craft Layer): Apply the system to layout, components, motion, iconography, and accessibility.
    4. Phase 3 (Documentation): Write or update a DESIGN.md file at the project root to serve as the single source of truth.
  7. Analyze stakeholder motivation using MICE

    main

    For every stakeholder involved in a negotiation, infer their primary motivator using the MICE framework to tailor your communication:

    • M (Money): Driven by bonuses, commissions, P&L responsibility, or compensation tied to the outcome.
    • I (Ideology): Driven by personal belief in a mission, category, or cause.
    • C (Compromise): Driven by consensus-building. Common in HR, procurement, or hiring committees. Strategy: Make their job of building consensus easier; do not focus solely on technical excellence.
    • E (Ego): Driven by visibility and reputation. Strategy: Provide them with visibility; underplaying them ruins the relationship.
  8. Reference materials for Technical Article Writer phases

    main

    The technical-article-writer skill utilizes specific reference files to deepen the quality of certain workflow phases. If you are implementing or extending this skill, refer to these files for the required logic:

    • Phase 2 (Title generation) & Phase 3 (Hook and intro): Use references/hooks-and-titles.md. This contains the 10 hook types, 6 copywriting frameworks (PAS, AIDA, BAB, FAB, PASTOR, 4Us), headline formulas, and research data.
    • Phase 4 (Body structure): Use references/article-structures.md. This contains detailed templates for 8 content types, the Diataxis framework, structural anti-patterns, and transition techniques.
  9. Understand the Three-layer token discipline

    main

    To maintain a scalable and maintainable design system across all stacks, follow this three-layer hierarchy:

    1. Primitives (raw): The base values (e.g., --blue-9, --space-3, --font-mono).
    2. Semantic (purpose): Tokens that describe how a value is used (e.g., --color-bg, --color-fg, --color-accent, --color-border).
    3. Component (variant): Tokens specific to a UI element (e.g., --button-bg, --card-radius).

    Crucial Rule: Components must reference semantic tokens, never raw primitives. This allows you to redefine semantic variables (e.g., under a .dark class) to update the entire UI without changing component logic.

  10. Strip AI discourse-architecture patterns

    main

    Even if individual words are compliant, the structural organization of a passage can signal AI generation. Avoid these structural patterns in descriptive or explanatory STE text:

    • Redundant Recaps: Do not use closing paragraphs that restate the opening. End where the last necessary instruction ends.
    • Catalog Structure without Throughline: Avoid covering topics at equal, shallow depth. Ensure sections depend on each other or are operationally relevant.
    • False Balance: Do not pair every fact with a counterbalancing qualifier (e.g., "However, it should also be noted..."). State the actual condition (e.g., "It depends on the operating pressure").
    • Ghost Q&A: Avoid self-directed questions (e.g., "Why is this step necessary?"). State the fact directly.
    • Constant Granularity: Real technical writing must "change altitude" by dropping into specific, verifiable details (torque values, error codes, part numbers). If data is missing, flag it as non-compliant rather than using vague language.
    • Lack of Scope: Always state what the document applies to. Avoid generic instructions; specify the model, system, or version (e.g., "On the X200 series pump...").
    • List as a Substitute for a Decision: Do not use a bulleted list to avoid choosing a specific value. State the correct answer directly.
  11. Use the CRAP checklist for layout design

    main

    The CRAP principles (Contrast, Repetition, Alignment, Proximity) serve as a portable checklist to prevent 'slop' (weak contrast, arbitrary placement, and uniform spacing).

    • Contrast: If two elements are not the same, make them significantly different. Avoid 'timid contrast' (e.g., 16px vs 18px or very similar grays).
    • Repetition: Repeat visual motifs (radius, spacing units, accents, type treatments) to unify the UI and teach the user the system.
    • Alignment: Ensure every element aligns to something. Strong left or grid alignment is generally superior to centering everything.
    • Proximity: Use space to group related items and separate unrelated ones.