| 13 | Introduction accuracy | Verify the highlights summary in the intro matches the actual content below. If it doesn't, rewrite the highlights to reflect the 2-3 most impactful items in the body |
| 14 | Enterprise context check | Every major bullet should have a "so what" phrase explaining enterprise relevance. Flag items that are purely descriptive with no enterprise value statement |
| 15 | Link text accuracy | Verify link labels match the source type: [Changelog] for changelog entries, [Announcement] for blog posts, [Release Notes] for release pages, [Docs] for documentation |
| 16 | Wording precision scan | Flag and review: e.g. vs i.e. usage, passive constructions, redundant qualifiers, pronoun ambiguity |
| 17 | Superlative audit | Flag universal quantifiers (any, every, all) applied to platform capabilities. Replace with specific or comparative framing (more, specific counts, expanding). "any agent" becomes "more agents". "all models" becomes specific model names |
| 18 | Surface attribution audit | For each bullet, verify the feature belongs in the section it is placed in. github.com features go under Platform/github.com sections, NOT under an IDE section. Common misattributions: Agents Tab = github.com (not VS Code), ACP = CLI (not VS Code). Check the feature's actual URL/docs to confirm |
| 19 | Cross-section label consistency audit (L67) | Scan for every model name and feature name that appears in MORE THAN ONE section (e.g., model bullet + IDE parity, or lead section + Latest Releases). Verify the GA/PREVIEW label is IDENTICAL in each occurrence. If conflicting, use the label from the official changelog announcement. Common conflict: model extensions to new IDEs get labeled GA in IDE parity but PREVIEW in model bullet. |
| 20 | Risky claim source audit | Billing, pricing, model availability, legal/preview, CCC/DPA, BYOK, OpenTelemetry, and cost-optimization claims require current source support. Use first-party GitHub/Microsoft sources for Copilot product and billing behavior; provider docs may support only general token-economics context. |
| 21 | Evidence-label audit | Optimization claims must label the evidence type: official behavior, GitHub usage report/export, internal telemetry, workflow experiment, anecdote, recommendation, or inference. Remove or hedge any claim that implies billing proof, durable savings, model superiority, or fleet readiness without evidence. |
| 22 | Action-owner audit | Each major section should include a recommended action, owner cue, or explicit awareness only treatment. This is mandatory for lead sections, billing/governance sections, legal/preview notes, and high-risk migration/deprecation content. |
| 23 | BYOK scope audit | Replace generic BYOK token usage wording with the specific surface and billing/reporting context: VS Code BYOK, Copilot CLI external/local provider, SDK/custom workflow, direct provider billing, model availability, or GitHub AI Credits impact. |
| 24 | OpenTelemetry scope audit | Scope OpenTelemetry to the actual instrumented surface. For SDK-backed workflows, cite Copilot SDK/OpenTelemetry. For standard VS Code or Copilot CLI usage, prefer GitHub usage reports, dashboards, and available IDE/CLI telemetry unless the source explicitly documents traces. |
| 25 | Developer billing caveat audit | When UBB or AI Credits affect developer behavior, confirm whether the copy needs the code-completions/Next Edit Suggestions caveat and Copilot code review caveat. Code review may involve AI Credits and GitHub Actions minutes; completions/NES are outside AI Credit billing for paid plans. |
