Systems Lab

Agent skill

content-eval

activeReaches the webActs undeclared1,220 words

Filed under Content and SEO.

From ericosiu/ai-marketing-skills · 21 skills · 3,449 · pushed 2026-08-16

What it does when it runs

Read from the skill and the 6 files bundled beside it. A skill’s own description is written to be selected by an agent, so it describes the job and not the dependencies.

Keys and connectors you must supply
None found.
Hosts it reaches
  • levelingup.beehiiv.com
  • singlebrain.com
  • www.singlegrain.com
Tool permissions it declares
No allowed-tools in the frontmatter. It does act, so it runs under whatever permissions your session already grants.
Actions present in the files
shell

Ask about content-eval

Opens your assistant with this page's verified links already in the prompt.

Is this safe to install?ClaudeChatGPT
Adapt it to my stackClaudeChatGPT
What else do I need for it to workClaudeChatGPT
Rather ask a human? Talk to Cheetah
git clone --depth 1 --filter=blob:none --sparse https://github.com/ericosiu/ai-marketing-skills.git /tmp/ai-marketing-skills
git -C /tmp/ai-marketing-skills sparse-checkout set "content-eval"
mkdir -p ~/.claude/skills/content-eval
cp -R "/tmp/ai-marketing-skills/content-eval/." ~/.claude/skills/content-eval/

Picked up without a restart. A project skill of the same name is shadowed by your personal one. For one repository only, swap ~/.claude/skills for .claude/skills. Claude Code docs ↗

The folder is the same in every client that implements the format — 46 of them — so if yours is not above, only the destination changes.

Reproduced in full from ericosiu/ai-marketing-skills/blob/2eb0f34edb8d6111ca8b2930fed92413c9af7002/content-eval/SKILL.md, which is licensed MIT (repository). 1,220 words, 26 headings.

name: content-eval description: >- Generate and score content ideas using an expert panel. Pulls from podcast transcripts, meeting notes, competitor analysis, and trending topics to produce a ranked content menu with production schedule. Use when asked to: "content eval", "score content ideas", "weekly content menu", "what should I film", "content ideas", "rate these video ideas".

Content Eval

Content ideation + expert panel scoring pipeline. Ingests raw material, generates ideas across your messaging pillars, scores them via a 7-expert panel, and outputs a ranked list with production schedule.


Step 1: Gather Raw Material

Collect signal from all available sources. Skip any source that's unavailable.

Podcast episodes

  • Read recent episodes from your podcast transcript directory (last 7 days)
  • Extract: topics covered, guest insights, audience questions, contrarian takes
  • Note episode titles for dedup against new ideas

Meeting notes

  • Check your meeting notes directory for recent notes
  • Extract: client questions, recurring themes, interesting moments, pain points
  • Focus on what your target buyers are actually asking about

Sales call insights

  • Check your call recording platform data for recent calls
  • Extract: objection patterns, recurring questions, competitor mentions
  • Note what prospects are confused about or struggling with

Trending topics

  • Note any topics the user mentions directly
  • Check competitor scan results (Step 2) for trending formats/topics
  • Look for news hooks or industry shifts you could react to

Step 2: Competitive Scan (if not provided)

If the user hasn't supplied competitive data, run a scan using your YouTube competitive analysis tool.

Channel sets are defined in references/competitors.md (same directory as this skill).

Analyze results

  • Outlier videos: Identify any video with >2x the channel's average views
  • Topic patterns: What subjects are multiple creators covering?
  • Format patterns: What formats are performing (listicles, reactions, tutorials, story-driven)?
  • Content gaps: Topics where NO competitor is covering something you could own
    • Especially gaps at the intersection of your unique expertise areas
    • Or gaps where you have unique data (revenue numbers, internal metrics, tool costs)

If your competitive analysis tool is not available, skip this step and note it was skipped.


Step 3: Generate Ideas

Generate 20-30 content ideas across three formats. Read references/pillars.md for pillar definitions and references/voice-rules.md for content voice rules.

Format targets

FormatCountDetails
YouTube Long-form (10-20 min)8-10 ideasDeep dives, screen recordings, frameworks
YouTube Shorts (<60 sec)8-10 ideasOne punch, one insight, one hook
X / LinkedIn Articles5-7 ideasManifesto-style, data-heavy, contrarian takes

Pillar requirement

Every idea MUST connect to at least one pillar defined in references/pillars.md. Ideas that don't clearly serve a pillar get killed.

Idea format

Each idea needs:

  • Title — specific, hook-driven, follows voice-rules.md
  • Description — 1-2 sentences on the content and angle
  • Format — Long-form / Short / Article
  • Pillar(s) — which pillar(s) it serves
  • Source signal — what raw material inspired it (podcast topic, competitor gap, client question, etc.)

Dedup check

  • Check recent published content (last 30 days)
  • If a similar angle was covered recently, either kill it or document a genuinely new hook
  • Apply the dedup rule from references/voice-rules.md

Manual override

If the user passes specific ideas to score (e.g., "score these content ideas: [list]"), skip idea generation and go directly to Step 4 with the provided ideas.


Step 4: Expert Panel Scoring

Run each idea through the 7-expert panel defined in references/panel.md.

Panel summary

#ExpertLens
1Viral Hook ExpertCuriosity gap, scroll-stopping power, title strength
2Algorithm ExpertCTR potential, watch time, search demand, recommendation likelihood
3Founder Brand StrategistPillar alignment, authenticity to founder's voice/experience
4B2B Buyer Persona ExpertWould a CMO/VP/CEO watch and want what you're selling?
5Content Differentiation ExpertIs anyone else making this? Unique angle?
6Short-form Adaptation ExpertCan it clip into 3+ viral Shorts? Quotable moments?
7Debate/Engagement ExpertComments, shares, disagreements potential

Scoring rules

  • Each expert scores 0-100
  • Pass threshold: 85+ average
  • For each idea, output:
    • Average score across all 7 experts
    • PASS / FAIL
    • Top expert comment (the most insightful feedback)
    • Biggest weakness (the lowest-scoring dimension and why)

Output table per idea

### [Idea Title] — [Format]
**Pillar(s):** [pillar names]
**Description:** [1-2 sentences]

| Expert | Score | Comment |
|--------|-------|---------|
| Viral Hook | [score] | [one-line] |
| Algorithm | [score] | [one-line] |
| Brand Strategist | [score] | [one-line] |
| B2B Buyer | [score] | [one-line] |
| Differentiation | [score] | [one-line] |
| Short-form | [score] | [one-line] |
| Debate/Engagement | [score] | [one-line] |

**Average: [score] — [PASS / FAIL]**
**Top insight:** [best expert comment]
**Biggest weakness:** [lowest dimension + why]

Scoring integrity

  • Be brutally honest. No grade inflation.
  • Generic ideas that anyone could make should score <70 on Differentiation.
  • Ideas without personal receipts/data should score <75 on Brand Strategist.
  • If an idea is genuinely great, let it score high. The threshold exists so only the best survive.

Step 5: Rank and Schedule

Rank passing ideas

Sort all ideas scoring 85+ by average score, highest first.

Create 4-week production schedule

Assign passing ideas to weeks based on effort vs impact:

WeekFocusTypical content
Week 1Low effort, high impactShorts that can ship same day, reaction clips
Week 2Medium effort, highest ceilingLong-form with screen recordings, tutorials
Week 3High effort, strategic anchorsArticles, manifesto pieces, deep-dive frameworks
Week 4Compounding contentReference pieces, reaction content, series starters

Kill list

List all ideas that scored <85 with:

  • Title
  • Average score
  • Primary reason for failure (the expert dimension that killed it)

Step 6: Output

Create a document with the full results.

Document structure:

  1. Executive Summary — Total ideas generated, pass rate, top 5
  2. Competitive Gaps Found — What no one else is covering
  3. Ranked Ideas (Passing) — Full scoring tables for each
  4. 4-Week Production Schedule — Visual calendar layout
  5. Kill List — Failed ideas with reasons
  6. Raw Material Summary — Sources used (podcasts, meetings, calls, trends)

Optionally post a summary to your team channel with:

  • Top 5 ideas (title, score, format, pillar)
  • Competitive gaps found (2-3 bullets)
  • Week 1 production schedule (what to film/write this week)
  • Link to the full document

Reference Files

FilePurposeWhen to read
references/pillars.mdMessaging pillar definitionsStep 3 (idea generation)
references/panel.md7-expert panel with scoring criteriaStep 4 (scoring)
references/competitors.mdYouTube competitor channel setsStep 2 (competitive scan)
references/voice-rules.mdContent voice/style rulesStep 3 (idea generation)

Files bundled with it

These load only when the skill asks for them, so they cost nothing until it runs.

Other skills for the same job

Different authors, same problem. Matched on the words in the skill name, across every library in the catalogue except this one.

Need help setting it up?

This page tells you what content-eval does and what it needs. Cheetah builds the agent setup it runs inside: data, CRM, sequencing and the guardrails.

Book a call →

The directory stays free. There is nothing gated behind this.