Content Quality: Clear Briefing (709 words by the script and my own count, inside the 300-800 Briefing range, headings included). Structure: Overview, Usage by client, Named skills/agents/MCP servers, Caveats, Access, What We Don't Know. Dense but accurate; every section maps to a changelog entry.
Source Verification: Read all three snapshots from disk (gunzip; text extracted from raw HTML) in sources/2026-10/github-copilot-usage-metrics-api-adds-per-client-totals-covering-opencode-and-claude-code-and-named-skills-custom-agents-and-mcp-servers/. source-0 (github.blog Oct 23 'Usage metrics API adds usage from tools like OpenCode', dated October 23, 2026): confirms totals_by_client array; scope (enterprise, organization, enterprise-user, organization-user, 1-day and 28-day); client field 'such as opencode, claude_code, codex, zed, copilot_cli, github_app, cloud_agent, code_review, vscode, or jetbrains' (an example list, 'such as', not exhaustive; unrecognized grouped under other); request_count 'The number of successful model requests from the client'; session_count; token_usage (prompt, output, average per request); totals_by_model with model_selection_mode user_selected/auto/product_selected; verbatim quotes 'Because those tools often don't send editor telemetry, their usage was missing from totals_by_ide and hard to account for.', 'counts jobs started by agent apps such as Claude and Codex on GitHub', 'Treat client as a best-effort grouping, not a verified identity.', 'request_count counts model calls, not prompts', 'Existing fields are unchanged, and totals_by_client is an empty array when there is no usage to report.'; cloud agent/code review product_selected and code review models unknown; access roles and policy requirement. source-1 (github.blog Oct 22 'Custom skills, agents, and MCP servers in usage metrics API', dated October 22, 2026): confirms names of skills, custom agents, MCP servers, slash commands and plugins instead of other/custom; totals_by_skill/custom_agent/mcp/slash_cmd/plugin; up to 25 entries per record up from five with trailing other; interaction_count and distinct_*_use_count unchanged; builds on September agentic CLI customization metrics; coverage list (Copilot CLI interactive and prompt mode, Copilot app, ACP integrations such as JetBrains, Copilot SDK); CLI 1.0.97 and SDK v1.0.21 minimums; quote 'Names aren't reported from an IDE's built-in chat telemetry'; 128 characters; file-path-like names stay other/custom; 'Review names before sharing reports outside your organization'; undercount quote 'These metrics come from client telemetry and can undercount activity, for example when clients are outdated or telemetry is blocked'; no backfill ('Names appear starting with report days processed after this release. Earlier days still show other or custom'). Attribution of entries to dates is correct, and the title is correct: per-client totals (OpenCode, Claude Code) are Oct 23; named skills/agents/MCP are Oct 22. The undercount and no-backfill caveats are in the Oct 22 entry only (article attributes them correctly); the best-effort client-label caveat is in the Oct 23 entry only (also correct). Both entries say availability is for enterprise owners, billing managers, organization owners and custom-role holders with View Copilot Metrics, policy must be enabled; neither labels the change preview or GA, and the article does not claim either. source-2 (docs.github.com REST page, API version 2026-03-10): is an endpoint index; it states 'the Copilot usage metrics policy must be set to Enabled everywhere for the enterprise', as the article says. It does NOT mention totals_by_client, OpenCode or Claude Code (grep on the raw HTML, snapshot and live: 0 hits); the article cites the page only for the policy requirement, which is supported. The docs page is live: snapshot fetched 2026-10-26T09:47Z; a live re-fetch by curl returned identical byte sizes for all three pages (107700, 107127, 453772), so no drift. Manifest suspicious_patterns null for all three; no injection text found. Hosts github.blog (allowlist line 889) and docs.github.com (line 2650) are on config/source_allowlist.txt by exact hostname; no allowlist edit needed. Three sources, all first-party GitHub: no independent corroboration, which is acceptable for a Briefing about GitHub's own product change.
Factual Accuracy: All numbers, field names, dates, version minimums and the 10 example client names match the raw snapshots; all six direct quotes checked verbatim. Minor wording note: the article says 'Both entries say the reports are available to...' which is accurate. The article says 'GitHub lists ... as examples', matching the 'such as' wording. Nothing unsourced; no orphan URLs in the body (3 body URLs = 3 sources).
Overall Assessment: Accurate, well-sourced Briefing; all claims, quotes and caveats trace to the two changelog entries and the docs page, with no corrections needed. Featured candidacy: 1 - accurate API-change coverage for admins and developers who use Copilot metrics, narrow audience