AI Skill Report Card

Governing Documentation

A-83·Sep 27, 2026·Source: Web

Documentation Governor

13 / 15

Given a project or product, produce four artifacts in this order:

  1. Documentation Tree — hierarchical map of all doc categories and files
  2. Version Strategy — how docs are versioned, tagged, and deprecated
  3. Knowledge Structure — ownership, cross-linking, and single-source-of-truth rules
  4. Maintenance Plan — cadence, triggers, and responsibilities for keeping docs current

Example invocation: "Set up documentation governance for our API platform with public docs, internal runbooks, and architecture decision records."

Recommendation▾
Add a third example showing a bad/anti-pattern output (e.g., team-organized tree) to reinforce contrast, since current examples only show good outcomes
14 / 15

Progress:

  • Step 1: Inventory existing documentation (types, locations, owners, last-updated dates)
  • Step 2: Identify audiences (external users, internal engineers, new hires, stakeholders)
  • Step 3: Build the Documentation Tree grouped by audience/purpose, not by team
  • Step 4: Define Version Strategy (semantic vs. date-based, deprecation policy, changelog rules)
  • Step 5: Define Knowledge Structure (canonical sources, cross-reference rules, naming conventions)
  • Step 6: Define Maintenance Plan (review cadence, staleness triggers, ownership matrix)
  • Step 7: Flag consistency risks (duplicated content, conflicting terminology, orphaned pages)

Step 3 details — Documentation Tree

Structure as a nested outline, max 3-4 levels deep:

/docs
  /getting-started
  /architecture
    /decision-records (ADR-XXX format)
    /diagrams
  /api-reference (auto-generated, versioned)
  /runbooks
  /governance (this doc's own meta-docs: style guide, templates)

Mark each node with: owner, audience, update trigger.

Step 4 details — Version Strategy

Recommend by default:

  • Semantic versioning (vMAJOR.MINOR) for anything tied to a shipped product/API
  • Date-stamped (YYYY-MM-DD) for runbooks, meeting notes, and living guides
  • Every doc carries a header: Version | Last Reviewed | Status (Draft/Active/Deprecated)
  • Deprecated docs redirect to successor, never deleted outright for 1 release cycle

Step 5 details — Knowledge Structure

  • One canonical source per concept; all other mentions link back, never duplicate
  • Naming convention: [Category]-[Topic]-[Subtopic].md
  • Cross-link using relative paths, validate links on each review cycle
  • Terminology glossary is mandatory and referenced, not restated, elsewhere

Step 6 details — Maintenance Plan

  • Review cadence: high-churn docs (API reference) reviewed per release; low-churn (architecture) reviewed quarterly
  • Staleness trigger: auto-flag any doc unreviewed for >90 days
  • Ownership matrix: each doc node has exactly one accountable owner (not a team)
  • Broken-link and orphan-page audit scheduled monthly
Recommendation▾
Include a concrete template snippet (e.g., full doc header format or ownership matrix table) rather than just describing it prose-style
15 / 20

Example 1: Input: "We have scattered README files, a Notion wiki, and Slack pinned messages for a mid-size SaaS product." Output:

  • Documentation Tree: consolidates into /docs with nodes for product-overview, api-reference, internal-runbooks, architecture-decisions, onboarding
  • Version Strategy: API reference gets semver tied to release tags; runbooks get date-stamped headers
  • Knowledge Structure: Notion wiki becomes canonical for onboarding only; Slack messages get migrated or archived with no standing authority
  • Maintenance Plan: monthly staleness audit, owner assigned per node, quarterly architecture review

Example 2: Input: "Single technical writer, small startup, just needs a lightweight structure before it gets messy." Output:

  • Documentation Tree: flat 2-level structure (/docs/product, /docs/internal)
  • Version Strategy: date-stamped only, no semver until first public release
  • Knowledge Structure: single glossary file, all docs link to it
  • Maintenance Plan: lightweight — writer self-reviews monthly, no formal ownership matrix needed yet
Recommendation▾
Address edge cases like multi-team ownership conflicts or migrating legacy docs with no clear owner
  • Always separate audience-facing docs from internal-only docs at the top tree level
  • Enforce "one canonical source" ruthlessly — duplication is the primary cause of drift
  • Tie version strategy to real triggers (releases, decisions) not arbitrary calendar dates alone
  • Keep the maintenance plan proportional to team size — don't impose enterprise process on a 3-person team
  • Always output a glossary/terminology node; inconsistent terms are the most common consistency failure
  • Don't organize the tree by internal team structure — organize by audience and purpose
  • Don't recommend deleting deprecated docs immediately — redirect and sunset instead
  • Don't propose the same review cadence for all doc types — high-churn and low-churn docs need different cycles
  • Don't skip the ownership matrix — "owned by the team" always decays into owned by no one
  • Don't over-engineer versioning for early-stage projects with no external consumers yet
0
Grade A-AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
13/15
Workflow
14/15
Examples
15/20
Completeness
17/20
Format
15/15
Conciseness
13/15