feat: improve skill scores for ai-marketing-claude

[rohan-tessl]

Hey @zubair-trabzada 👋

I ran your skills through tessl skill review at work and found some targeted improvements.

Here's the full before/after:

Skill	Before	After	Change
market-seo	46%	89%	+43
market-proposal	46%	89%	+43
market-funnel	48%	89%	+41
market-brand	43%	84%	+41
market	45%	85%	+40
market-audit	55%	89%	+34
market-landing	59%	89%	+30
market-social	59%	89%	+30
market-ads	60%	89%	+29
market-launch	61%	89%	+28
market-report-pdf	64%	89%	+25
market-competitors	60%	81%	+21
market-copy	69%	89%	+20
market-report	69%	89%	+20
market-emails	73%	89%	+16

Average: 57% → 88% (+31)

Changes made

All 15 skills

• Added YAML frontmatter — none of the skills had frontmatter (name, description), which caused them to fail validation entirely (scoring 0% without it). After adding frontmatter with proper kebab-case names and quoted descriptions, the LLM judge could actually evaluate the content.
• Added "Use when..." clauses — every description now includes explicit trigger guidance so agents know when to select each skill, which was the single biggest scoring improvement across the board.
• Expanded trigger terms — added natural user language variations (e.g., "GTM" for market-launch, "cart abandonment" for market-funnel, platform names for market-ads/market-social).

High-impact content improvements

• market-brand (43→84): Reduced from 477 to 180 lines (62%). Removed verbose tone signal tables and archetype descriptions Claude already knows. Added validation checkpoint for source material sufficiency.
• market-seo (46→89): Reduced from 488 to 128 lines (74%). Removed basic SEO concept explanations. Added script execution validation steps.
• market-proposal (46→89): Reduced from 387 to 158 lines (59%). Removed pricing psychology tips, follow-up sequences, and formatting guidance. Added input validation checkpoint.
• market-funnel (48→89): Reduced from 449 to 291 lines (35%). Removed funnel type reference tables and drop-off cause matrices. Added WebFetch example and validation checkpoints.
• market (45→85): Added concrete WebFetch example, replaced Business Context Detection section with Subagent Validation Checkpoints.

Moderate content improvements

• market-ads (60→89): Removed ROAS/CPA benchmark tables, condensed ad copy angle templates.
• market-copy (69→89): Removed CTA Color Psychology section, condensed page-specific copy structures.
• market-launch (61→89): Removed Common Launch Mistakes and Key Principles sections.
• market-social (59→89): Removed platform demographics tables, added brand voice validation checkpoint.
• market-landing (59→89): Condensed benchmark tables, hero scoring rubrics, and A/B test examples.
• market-competitors (60→81): Condensed discovery methods and template tables.
• market-audit (55→89): Condensed business type detection to bullet list format.

Lighter touch (description-only improvements)

• market-emails (73→89): Removed basic concept explanations and "Why It Matters" column.
• market-report (69→89): Removed E-E-A-T explanation and glossary section.
• market-report-pdf (64→89): Removed PDF vs Markdown comparison table and color scheme tables.

Note: All skills originally scored 0% because they lacked YAML frontmatter. I added frontmatter with proper names and descriptions to unblock the LLM judge, then re-scored to get meaningful baselines. The "before" scores in the table above reflect those post-frontmatter baselines — the true starting point was 0% across the board.

Honest disclosure — I work at @tesslio where we build tooling around skills like these. Not a pitch - just saw room for improvement and wanted to contribute.

Want to self-improve your skills? Just point your agent (Claude Code, Codex, etc.) at this Tessl guide and ask it to optimize your skill. Ping me - @rohan-tessl - if you hit any snags.

Thanks in advance 🙏