Kai Weekly Audit

/kai-weekly-audit — what changed this week, and what someone does about it Monday

Kai root note: knowledge/, harness/, and scripts/ paths in this skill live in the Kai install, not the user's project. Resolve them against the first ancestor directory of this SKILL.md that contains a knowledge/ folder (the Kai plugin root, ~/.claude/kai, or the kai-cmo-harness repo). MARKETING.md, memory/, and any output files live in the current project. If a referenced scripts/ command is not available in this install, say so, skip it, and continue with the file-based guidance — never fabricate its output.

Objective

A sourced weekly operating review: the last 7 complete days measured against the prior 7, every material change scored, and a short action list with owners. This is the cadence layer above /kai-audit, /kai-seo-audit, /kai-cro, /kai-daily-ad-review, /content-report, and watcher output — it does not re-run those audits, it reads their output and decides what needs a decision this week.

A review that scores a metric it cannot source is worse than a review that reports a gap.

Done when

Work type audit-report (also_covers: weekly-audit) — floor E3/C4/O1 (harness/eco-floors.yaml).

• E3 — a named human approved the exact delivered file, and every quantitative claim resolves to a collector artifact in the week's audit folder.
• C4 — the Kai Data Provenance Rule: collector run before writing, mode declared, a source cited for every number, missing sources written to _data-gaps.md, then python scripts/quality_gates/audit_provenance_lint.py workspace/audits/weekly/<YYYY-MM-DD> --audit-dir passes alongside banned_word_check.
• O1 — every P0/P1 action names the metric it targets, its current value, the threshold that counts as fixed, and an owner. Read at 60 days: were the recommendations adopted and did the metric move?

Constraints

• Provenance is non-negotiable. Load harness/references/audit-data-provenance.md before writing findings. Declare the mode: sales_external for prospect or public-only reviews, onboarding_connected when GSC, GA4, GBP, ads, CRM, call tracking, or client exports are connected, internal_demo when sample data is used.
• Never publish a review count, ranking, traffic figure, conversion, call volume, ad metric, Core Web Vital, backlink, or revenue number without source, retrieval date, and artifact path. Missing sources become _data-gaps.md entries, never estimates.
• Only score metrics present in audit-data.json, connected exports, or raw pull artifacts. Everything else is scored Gray.
• Every scored item carries its provenance block:

  yamlCopy code
  claim: ""
  source_tier: connected | public_observed | user_provided | inferred | missing_data
  source_name: ""
  source_url: ""
  retrieved_at: ""
  evidence_artifact: ""
  score_eligible: true | false
  

• Connected collectors run only when access is confirmed. If a command cannot run because credentials are missing, record the exact missing source in _data-gaps.md rather than substituting another source.
• Comparison window is fixed: last 7 complete days against the prior 7 complete days. Partial days are excluded.
• This skill reads and reports. It does not change ad accounts, pages, or sends. Recommended fixes route to the skill that owns the change, which carries its own approval gate.

Revenue lens

Load the Revenue Lens section of harness/references/audit-data-provenance.md with the provenance rule.

• Open with the funnel, not activity. Traffic source → intent → lead → follow-up → booked/paid, one row per stage, before any channel or technical finding. Sessions and lead counts are inputs, never the headline.
• Rank gaps by the dollars each stage leaks. High-intent traffic with no lead path, leads with no follow-up, and follow-up with no booking outrank traffic volume.
• Technical setup is one line. Schema, llms.txt, AI crawler access and sitemap get a sourced pass/fail; they become findings only when a measured check fails.
• Unverified means check. Any claim or funnel stage not verified in this run is labeled check with what would verify it, is not scored, and is mirrored in _data-gaps.md.

Context

Review areas — cover each where data exists:

Scorecard scale: Green = on track · Yellow = needs attention this week · Red = immediate owner decision or fix · Gray = not scored, data missing.

Actions group into three: Do this week (P0/P1 with owner and due date), Watch next week (trend risk, insufficient evidence), Needs data (sources required before the next review). Route each fix:

Output goes to workspace/audits/weekly/<YYYY-MM-DD>/: _data-sources.md, _data-gaps.md, audit-data.json, kai-data.json, _weekly-scorecard.md, _weekly-findings.md, _weekly-actions.md, _skill-routing.md, and html-presentation/index.html when the user asks for a client-ready artifact (via /kai-html-presentation).

Escalate when

• The URL, data mode, connected sources, or target conversion event is unknown and the review cannot be sourced.
• A collector fails and the missing source would change a Red/Green call.
• A metric moved sharply and the available data cannot distinguish a real change from a tracking break.
• A finding implies spend, a live-channel change, or a client-facing claim that has not been approved.