Skill Hub — Foundation

Quality Rules

What makes a Skill Hub page good or bad. Use these rules when creating, reviewing, or editing skill pages.

Good page criteria

A good Skill Hub page must satisfy all of these:

  • Actionable: A reader can follow the working method and produce output tomorrow.
  • Situation-specific: It names real work situations, not abstract topics.
  • Question-driven: It provides specific diagnostic questions, not generic prompts.
  • Rule-based: It includes "If X, then Y" decision rules.
  • Template-equipped: It includes at least one copy-paste-ready template.
  • Agent-ready: It has an Agent Instructions section with explicit do's and don'ts.
  • Honest: It states limitations and verification status.
  • Linked: It connects to related skills and Atlas pages with valid local links.
  • Mistake-aware: It lists common mistakes with consequences.
  • Quality-checked: It includes a verifiable quality checklist for its own output.

Weak page warning signs

These are signals that a page needs rewriting:

  • The page starts with a definition of the concept.
  • The "When to use" section has only one vague bullet like "when data quality is important."
  • The working method has fewer than 5 steps or the steps are not sequential.
  • There are no decision rules.
  • The template is missing or is just a heading with no structure.
  • The agent instructions are missing or say "use your best judgment."
  • The page reads like a textbook chapter or a blog post.
  • The page contains motivational language like "unlock the power of data."
  • The page claims to be "comprehensive" or "complete."
  • All examples are generic ("a company," "a system") instead of specific ("SAP S/4 sales order creation," "MDG business partner workflow").

Anti-generic writing rules

These rules prevent the most common quality failure: generic framework summaries.

Do not write Write instead
"Data governance is important because..." "If ownership is unclear, produce an ownership matrix before proposing automation."
"Event-driven architecture provides loose coupling..." "Decide whether an event should exist, who owns it, what its contract is, and how failures are monitored."
"Requirements elicitation is the process of..." "Turn vague stakeholder complaints into requirements, assumptions, risks, and acceptance criteria."
"Stakeholder analysis helps identify key people..." "Map who has decision rights, who has information, who will resist, and who is missing from the room."
"A root cause is the fundamental reason..." "Separate symptom, root cause, and prevention gap. If you only fix the symptom, the defect recurs."
"Master data management ensures consistency..." "Decide which system creates the record, which system can change it, and how duplicates are detected."

Public-safety rules

Every Skill Hub page must be safe to publish:

  • No private material: Do not publish client names, internal project notes, proprietary code, or private correspondence.
  • No private paths: Do not reference internal file paths, server names, or network locations.
  • No copyrighted text: Do not copy passages from DAMA-DMBOK, BABOK, TOGAF, or other copyrighted frameworks. Paraphrase in your own words.
  • No fake authority: Do not claim official framework certification, endorsement, or authority.
  • No fake citations: Do not invent sources, page numbers, or edition references.
  • No vendor bias: Do not promote one vendor as the only solution unless the skill explicitly covers that vendor's product.
  • No invented data: Do not make up statistics, survey results, or case study details.

Agent-readiness rules

Every skill page must be usable by an AI agent:

  • The Agent Instructions section is mandatory.
  • Agent instructions must specify what context to gather before applying the skill.
  • Agent instructions must tell the agent what artifacts to produce.
  • Agent instructions must list things to avoid (generic language, fake certainty, skipping steps).
  • Agent instructions must explain how to handle missing information.
  • The working method must be sequential and unambiguous. An agent should not need to guess the order.
  • Decision rules must be explicit. "Use judgment" is not a decision rule.
  • Templates must be copy-paste ready with clear fill-in-the-blank fields.

Artifact-readiness rules

Every skill must produce usable artifacts:

  • At least one template must be included directly on the skill page.
  • Templates must have a clear artifact name and ID convention.
  • Templates must include all fields needed for the artifact to be useful.
  • Templates must distinguish between required and optional fields.
  • Deliverables must be named and described, not just listed as nouns.
  • Quality checklists must be verifiable yes/no items.

Link quality rules

Links must be valid and purposeful:

  • All local links must resolve to existing pages. Do not invent links.
  • Do not create broken links. Verify before committing.
  • Do not over-link. Two to five related skills is enough.
  • Only link to Atlas pages that add real diagnostic or conceptual value.
  • External links must have a clear purpose and be verifiable.
  • Do not link to every page in the same group. Link to the most relevant ones.

Review checklist for new skill pages

Before a skill page is accepted:

  • [ ] Page follows the Skill Page Template structure.
  • [ ] Page passes all good page criteria above.
  • [ ] Page has no weak page warning signs.
  • [ ] Page follows anti-generic writing rules.
  • [ ] Page passes public-safety rules.
  • [ ] Page passes agent-readiness rules.
  • [ ] Page passes artifact-readiness rules.
  • [ ] All local links are valid.
  • [ ] No private material is exposed.
  • [ ] No copyrighted text is copied.
  • [ ] Verification status and limitations are stated.