SKILL: Tessl Article Review — Editorial, Audience, SEO & AEO Scorecard

Step 0: Classify the Article Type

Classify the article into one of these types before scoring — this determines what "good" looks like.

State the type clearly at the top of your review. Also state the declared (or recommended) primary persona.

When the type doesn't match the format

Always frame type shifts as additive ("add X"), not subtractive ("this shouldn't be Y"). For example: a piece that "reads like docs but it's on the blog" needs "Add a 2-sentence problem statement before the first H2 explaining why someone would need this" — not "remove the technical detail."

Step 1: Preserve the Author's Voice

You are reviewing, not rewriting. The author's voice, style, and personality must be preserved.

Flag only these house style issues:

• Em dashes (—): flag every instance (house style: remove them)
• Declarative claims about third-party tools: make suggestive ("aims to" > "is built for")
• Vibe-coding references that aren't warranted by the topic
• Hype words ("revolutionary", "game-changing", "cutting-edge", "supercharge", "unlock")
• Generic AI preamble in opening paragraph ("In a world where AI moves faster than ever…")
• Factual inaccuracies or misleading framings

Everything else — dry humor, technical precision, conversational energy, long flowing paragraphs — is the author's call. Title suggestions and improvement recommendations must match the author's existing tone and energy.

Step 2: Verify Audience & Spectrum Anchoring

Before scoring, run two filters:

Persona declaration

Every article must declare a primary persona in its metadata. If absent, flag it and infer the best candidate from the framing.

• Economic Buyer — VP Eng / Director of Eng / CTO: cares about 2× developer productivity, fewer defects, faster delivery, governance, risk, predictable ROI. Reads at the trend and pattern level.
• Champion / Enabler — Head of Platform Eng / DevEx / AEL: cares about reliable agents in real workflows, scaling adoption, standards, integrations, KPIs that prove enablement impact.
• Coach / Agent Adoption Advocate — Senior Dev / Staff Eng / AI Engineer: use sparingly; technically credible, production-grade.
• User — Developer / Engineer: rarely the primary target now. Push to reframe for AEL or Buyer unless the topic is genuinely IC-only.

Default bias: VP Eng or AEL. Coach pieces are occasional. User pieces should be questioned.

Spectrum anchoring

• Multi-example Trend Analysis is the default and strongest format: 3-7 named tools/companies/reports showing a pattern.
• Single-launch pieces are acceptable when (a) the launch genuinely moves the strategic picture, (b) the article connects it to comparable tools and the wider category, and (c) the "so what for engineering leaders" is clear.
• Standalone vendor coverage with no broader frame is the failure mode.

Context: Who is Tessl?

• Agent enablement platform helping engineering orgs scale skills (and other context engineering) across coding agents (Claude Code, Cursor, Copilot, Devin, etc.).
• Skills: reusable markdown-based instruction sets for coding agents, installable and versioned like packages. The Tessl Registry holds 2,000+ evaluated skills (tessl search / tessl install). tessl.json tracks installed skills analogous to package.json.
• Evals: review evals (structure against best practices) and task evals (running agents with/without skills, measuring outcomes). Core value prop: "You can't optimize what you can't measure."
• Target audience: VP Eng / Director Eng / CTO (Economic Buyer) and AI Enablement Leads / Head of Platform Eng (Champion). ICP is ~50+ engineer orgs. They care about reliability, governance, measurable enablement, and scaling AI adoption — not hype.

Scoring Categories

Score each category 0–5. Total possible: 30. For each score level: 0–1 = absent/broken; 2 = partial/implicit; 3 = meets the bar; 4 = strong execution; 5 = exceptional/category-defining.

Category 1: Audience & Angle

Most strategic dimension under the current direction. If this fails, the piece doesn't ship as-is.

Category 2: SEO & Discoverability

Category 3: AEO/GEO Extractability & Originality

A growing share of readers will ask ChatGPT, Claude, Perplexity, or Gemini rather than visit tessl.io. The unit of optimization shifts from "rank in the top 10" to "be the source the model quotes."

What to check:

• Extractable answers throughout — can a 2-3 sentence chunk be lifted out as a coherent answer for every major claim?
• Question-shaped H2s/H3s — at least one heading maps to a natural-language query a VP Eng or AEL might ask an LLM.
• Front-loaded answers — every major section opens with a 1-2 sentence direct answer before elaboration.
• Definition format — "X is Y. It does Z. It matters because W." for at least one core concept.
• Named entities — 3-5+ specific tools, companies, reports, products.
• Dated claims — "as of 2026" where appropriate; avoid undated absolutes that age fast.
• Original / proprietary content — Tessl-specific data, branded frameworks, named quotes, eval results.
• FAQ block for Trend Analysis / Thought Leadership / Comparison pieces.
• Extractable list or table — at least one comparison table or enumerable list.
• No walls of text — no paragraph longer than ~5 sentences without sub-structure.

Category 4: Tessl Product & Strategic Alignment

Category 5: Technical Depth & Accuracy

Category 6: Structure & Readability

Output Format

The review is organised per strategic dimension, not per article section. For each of the six dimensions, give a 0-5 score AND atom-level recommendations grouped under that dimension. End with the SEO package and Tessl product touchpoint.

The format is designed to be pasted directly into the article's review page on Tessl.

Output template

📝 ARTICLE REVIEW — [Article Title]
📂 Type: [Thought leadership / Trend analysis / Skill showcase / Tutorial / News / Comparison]
👥 Primary persona: [VP Eng (Buyer) / AEL (Champion) / Coach / User]  (declared / inferred)
🌐 Spectrum anchoring: [Trend-led / Single-launch with framing / Standalone — flag]

OVERALL: [X]/30  →  Verdict: [Publish-ready / Strong draft / Needs work / Major revision / Start over]
Hard fails: [list each, or "None"]

═══════════════════════════════════════
1. AUDIENCE & ANGLE — [X]/5
═══════════════════════════════════════
✓ Working: [1-2 sentences on what's strong about how this piece serves its declared persona, spectrum framing, and "so what for engineering leaders"]
→ Recommendations:
  • [Atom-level fix, e.g., "Persona is declared as VP Eng but section 3 reads as IC-targeted — elevate the framing by replacing the four-primitives walkthrough with a 'why this matters for engineering leaders' paragraph"]
  • [Atom-level fix]

═══════════════════════════════════════
2. SEO & DISCOVERABILITY — [X]/5
═══════════════════════════════════════
✓ Working: [1-2 sentences on keyword discipline, H2 structure, internal links]
→ Recommendations:
  • [Atom-level fix, e.g., "Title is 81 chars; trim to ≤60 and lead with primary keyword: 'Agent Context Governance: Why Multi-Agent AI Breaks at Scale'"]
  • [Atom-level fix]

═══════════════════════════════════════
3. AEO/GEO & ORIGINALITY — [X]/5
═══════════════════════════════════════
✓ Working: [1-2 sentences on named entities, extractable atoms, original content]
→ Recommendations:
  • [Atom-level fix, e.g., "Add a definition block for 'state divergence' in section 2 — format as 'X is Y. It does Z. It matters because W.'"]
  • [Atom-level fix, e.g., "Add a 3-question FAQ at the end — 'What is shared agent memory?', 'How does it differ from a vector database?', 'What is state divergence?'"]

═══════════════════════════════════════
4. TESSL ALIGNMENT — [X]/5
═══════════════════════════════════════
✓ Working: [1-2 sentences on how the piece connects to Tessl's worldview]
→ Recommendations:
  • [Atom-level fix]

═══════════════════════════════════════
5. TECHNICAL DEPTH — [X]/5
═══════════════════════════════════════
✓ Working: [1-2 sentences on accuracy, depth, practical guidance]
→ Recommendations:
  • [Atom-level fix]

═══════════════════════════════════════
6. STRUCTURE & READABILITY — [X]/5
═══════════════════════════════════════
✓ Working: [1-2 sentences on flow, opening, closing, H2 hierarchy]
→ Recommendations:
  • [Atom-level fix, e.g., "Opening paragraph buries the lede; lead with the direct claim before going to the Ranganathan quote"]
  • [Atom-level fix]

═══════════════════════════════════════
SEO PACKAGE
═══════════════════════════════════════
Primary keyword: "[ONE keyword from Cluster 1-5 — commit to it]"
Why this keyword: [1 sentence naming the cluster and matching search intent]
Meta description (≤155 chars): "[the actual meta description]"
URL slug: /blog/[slug]
Title tweak: [exact suggested title, or "Current title works"]
H2 keyword opportunities: [before → after rewrites for any H2s that could carry searchable terms]
Internal links: [2-3 specific tessl.io URLs with the phrase in the article that should become the anchor]

═══════════════════════════════════════
TESSL PRODUCT TOUCHPOINT
═══════════════════════════════════════
[Where a natural reference to skills/evals/registry/CLI could be woven in. If already done well, say so.]

Format rules

• Scores live ONLY on the six strategic dimensions. Each is 0-5. Total is /30. Do NOT score article sections individually.
• Recommendations are grouped under the dimension they belong to. A single fix can be referenced from multiple dimensions if it spans them.
• Recommendations must be atom-level. "Improve the opening" is not acceptable. "Replace the em dash on line 1 with a comma" or "Lead with the direct claim before the Ranganathan quote" is. Quote phrases or reference exact heading text.
• The "✓ Working" line is 1-2 sentences per dimension. Tell the author what to preserve, not just what to fix.
• No "Google Docs comment" framing. The output is the review. Paste it directly into Tessl's article review interface.

SEO Package: How to Choose the Primary Keyword

Commit to ONE primary keyword — do not present options and let the author choose.

Decision process:

1. What is the single question this article answers?
2. Match that question to the closest keyword from the priority clusters below.
3. If the article spans multiple clusters, pick the one with the strongest match to the article's unique angle, not just its topic.
4. If no priority cluster fits, suggest a long-tail keyword and flag it — pieces outside the clusters should be questioned for priority.

When suggesting a title tweak: Match the author's existing tone (playful, technical, provocative, straightforward). Weave the primary keyword into the existing title structure — don't replace it. If the current title already works, say so.

Scoring Thresholds & Recommendations

Hard fails (override the score)

Regardless of total, the piece does not ship if any of these are true:

• Em dashes present (title or body). Easy to fix; non-negotiable.
• Hype words present ("revolutionary", "game-changing", "cutting-edge", "supercharge", "unlock"). Replace with specifics.
• Persona is User without strong justification — push to reframe for AEL or Buyer.
• No spectrum anchoring at all — standalone vendor coverage with no broader category framing.
• AEO/GEO score ≤1 — the piece needs a structural rewrite, not a polish pass.
• Marketing-shaped opening paragraph — generic AI preamble. Rewrite the opening.

Review Process

1. Classify the article type (Step 0). State it explicitly. Recommend a primary persona if one isn't declared.
2. Run the audience and spectrum filters (Step 2). Flag any gaps.
3. Note the author's voice (Step 1) — tone, style, personality. All suggestions must match it.
4. Read the article fully before scoring. Take mental notes on issues that span the piece (em dashes, persona drift, AEO gaps, weak openings, etc.).
5. Score each of the six strategic dimensions 0-5. For each dimension, write a 1-2 sentence "✓ Working" line and 2-4 atom-level recommendations. Quote specific phrases or reference exact heading text.
6. Generate the SEO package. Commit to ONE primary keyword from Cluster 1-5. Write the actual meta description and URL slug — don't punt this to the author.
7. Identify the Tessl product touchpoint. Where could skills/evals/registry/CLI be referenced naturally?
8. Produce the output using the per-dimension template above. Paste-ready for the article's review page on Tessl.
9. Be direct but constructive. The author is a colleague, not a student.

Reference: Tessl SEO Priority Keywords

Cluster 1: Agent Skills (High Priority)

agent skills, AI agent skills, coding agent skills, agent skill registry, agent skill evaluation, how to create agent skills, MCP skills, agent skill package manager

Cluster 2: Context Engineering (High Priority)

context engineering, context engineering AI agents, context engineering vs prompt engineering, AI agent context management, effective context engineering

Cluster 3: AI Agent Evaluation (High Priority)

AI agent evaluation, AI agent evals, how to evaluate AI agents, AI agent testing framework, agent evaluation metrics

Cluster 4: AI Agent Reliability (Mid Priority)

AI coding agent accuracy, improve AI agent accuracy, AI agent hallucination fix, coding agent reliability, reduce AI coding errors

Cluster 5: Context Governance & Security (High Priority)

context governance, context security, secure context engineering, AI context governance, agent context governance, AI agent governance, context risk management, secure AI context, context compliance, context engineering security, prompt injection enterprise, agent security posture, skill governance, skill security review, AI agent risk management, context audit

Decision-maker themes (agent enablement, productivity ROI, FinOps, harnesses) are not separate clusters. Weave them into pieces anchored on the five clusters above, particularly Clusters 1, 2, and 5.

Keyword Selection by Article Type

Reference: House Style Quick-Check

• No em dashes (—): flag every instance
• No hype words ("revolutionary", "game-changing", "cutting-edge", "supercharge", "unlock")
• Suggestive tone for third-party claims ("aims to" > "is built for")
• Minimal vibe-coding references (unless warranted)
• 2–5 H2 subheadings minimum
• At least 1 internal link to another tessl.io article (prefer 2+)
• At least one question-shaped H2 or H3
• No excessive bolding or bullet-point overuse in prose sections
• Opening paragraph front-loads the topic with specificity (no generic AI preamble)
• Closing is takeaway, question, or pointer — not a pitch
• Terminology: "skills" not "tiles" (unless article specifically explains the legacy term)
• Author's voice and personality are preserved throughout all suggestions
• Honest about uncertainty where the pattern is still forming

Reference: Persona × Article Type Fit