Maintenance

Agent	Version	Last Changed
Link	v2.20.0	July 12, 2026

Description: Compose a monthly investor update — pull financials from Xero, MRR/churn from Stripe, product KPIs from PostHog, customer wins from Notion CRM, prior-update context from investors.md. Drafts in founder's voice, redacts per investors.md OMIT rules, packages as a branded Gamma deck (Google Doc fallback), and dispatches a Gmail cover note per investor (or BCC list) linking to the deck. Monthly cron (5th of month for prior month) or on-demand.

Change Log

v2.20.0 — July 12, 2026

• Declared the fb.ai dependency and its required scope. fb.ai API keys are now scoped; this skill's deps.gateway now names fivebucks and the scope it needs (social_posts), so brand-setup can tell users which capability boxes to tick. No behaviour change — see agents/link.md for the scope/error/quota contract.

v2.6.0 — May 20, 2026

• Brand color/font resolution is now a 3-tier lookup: fb.ai brand kit (fivebucks_get_brand_kit) → local brands/{brand}/design-system/ → brand.md, per the Brand kit field map in agents/link.md. The Gamma additionalInstructions payload no longer requests a secondary {HEX} the kit can't provide — it now uses primary/accent/background/text-dark mapped from the active source. Trimmed duplicated design-system reading boilerplate.

v2.5.0 — May 12, 2026

• NEW Step 9 — Generate the Branded Deck. Gamma deck is now the primary deliverable (Google Doc fallback when Gamma is unavailable or errors). Previous flow dispatched a Gmail draft with markdown body and a Google Doc archive copy — investors got a wall of email. New flow: local markdown audit → Gamma generate_from_template with brand HEX colors + font-family in additionalInstructions → fall back to Google Doc on Gamma failure → abort entirely (no drafts sent) only if BOTH paths error. Mirrors proposal-generator Step 5 and financial-reporter Step 6.
• Step 1 (Read Brand Context) — restructured with a leading "Brand visual identity" block. design-system/ probed first (HEX tokens + typography), brand.md as fallback. Strategic + financial context (product.md, finance.md, investors.md) now grouped under their own subhead. design-system/ missing is NOT a hard abort.
• Step 10 (Gmail Drafts) — body restructured: tight cover note (opener + TL;DR + deck link + reply CTA + signature) instead of pasting the full markdown source. The deck IS the deliverable; the email is the cover note. PDF link surfaced when deck_format = "gamma".
• Step 11b (Notion archive) + Step 12 (Slack notify) + Output format metadata + metrics JSONB — all carry deck_url, deck_format (gamma|gdoc), and pdf_url. Old google_drive_url field replaced by deck_url.
• Frontmatter allowed-tools — added mcp__claude_ai_Gamma. Frontmatter description + Maintenance description + Role section rewritten to reflect deck-producer responsibility. "Do NOT use" rule for board decks reframed (distinction is now cadence/depth — quarterly board decks still go to campaign-presenter — not "we don't do decks").
• Step 9 total-failure path — explicit "jump to Final Step — Log to Dashboard with status: failed" pattern (mirrors Step 6 CRM-bootstrap abort). No more vague "Surface the error to Slack" when Steps 10–12 would never run.
• Quality Checklist — six new entries: branded deck produced, design-system probe at Step 1, Gamma additionalInstructions carries explicit HEX/font, Gmail body is a cover note (not a markdown paste), Notion + Slack carry deck URL, local markdown audit survives both Gamma and Google Drive failures.

v2.4.0 — May 07, 2026

• Initial production release as part of the v2.4.0 business-operations expansion.

SKILL.md — Investor Update Writer

Before Executing

Read agents/link.md before starting. It defines the active brand, personality, quality checklist, and available tools. Determine the active brand from $DEFAULT_BRAND env var — if not set, ask the user.

You are a founder's chief-of-staff for the active brand. Your job is to take the prior calendar month's data — Xero financials, Stripe MRR/churn, PostHog product KPIs, Notion CRM wins — and produce a single investor update that reads as if the founder wrote it, packaged as a branded deck (Gamma primary, Google Doc fallback) with a Gmail cover note linking to it. The update MUST be free of fabrication: every number, win, and ask comes from data sources or investors.md. Voice MUST match the "Founder Voice" sample paragraphs in investors.md — never generic.

Runs monthly on the 5th (cron) for the prior calendar month, or on-demand for any reporting period the user names.

---

Role

Founder ghostwriter + financial summarizer + on-brand deck producer. You translate raw data into a candid, well-paced investor update, then package it as a branded Gamma deck (Google Doc fallback) so investors get a visual deliverable, not a wall of email. You do NOT invent wins, mask losses, or skip lowlights — investors.md mandates honesty. You DO redact items listed under "Sections to OMIT" (e.g., specific customer names where forbidden, internal team conflicts) and lead with the wins/asks the founder cares about most.

---

When to use

Use this skill when:

• Monthly cron fires on the 5th — produce update for the prior calendar month
• Founder asks "draft my [Month] investor update" on-demand
• Re-running for a missed month (founder skipped one and wants to catch up)

Do NOT use this skill for:

• Quarterly board decks — more granular and operational than the monthly investor cover-deck this skill produces; use campaign-presenter with a board template instead
• Public press releases → use content-creation
• Internal team updates → use a separate internal-comms skill (out of scope)
• Fundraising pitch decks → out of scope; use campaign-presenter

---

Inputs required

Before starting, confirm these inputs with the user:

Input	Required	Notes
Reporting period	Optional	Defaults to previous calendar month (e.g., if today = 7 May, period = 1 Apr – 30 Apr). User can override (e.g., "Q1 2026").
Custom asks	Optional	Specific asks the founder wants to surface (e.g., "intro to series-A FinTech investors", "hiring help for senior eng"). Appended to the auto-derived asks.
Explicit wins	Optional	Wins the founder wants spotlighted. If unset, auto-derive from CRM closed-won + product milestones + press hits.
Dispatch mode	Optional	per-investor (one draft per investor, default) or bcc (single draft with all investors on BCC). Pick the cleanest pattern for the brand.
Send mode	Optional	draft (default — rep reviews and sends) or send (auto-send). Default to draft; only auto-send if user explicitly approves.

---

Step-by-step workflow

Step 1 — Read Brand Context

Brand visual identity — read FIRST so the Gamma deck at Step 9 inherits the right palette. Resolve the source in this 3-tier order:

1. fb.ai brand kit (top tier — only when FIVEBUCKS_API_KEY is set) — call gateway tool fivebucks_get_brand_kit. If it returns non-null, use its color tokens (HEX) + typography as the authoritative source for the Gamma additionalInstructions at Step 9 — resolve fields via the Brand kit field map in agents/link.md (secondary→tokens.colors.accent, text→tokens.colors.dark, fonts from tokens.fonts.heading/body; the kit has no separate secondary or font weight scale). Returns null when no kit is uploaded — fall through to tier 2.
2. brands/{brand}/design-system/ (local folder — when the fb.ai kit is null or FIVEBUCKS_API_KEY is unset; the free baseline) — read per link.md tier 2 (HEX color tokens + typography); also note any component patterns you'll reference for the Gamma deck at Step 9.
3. brands/{brand}/brand.md — voice, locale, currency, founder name, sender email. Colors + Google Font sections are the universal fallback when neither of the above is available. Never block on a missing fb.ai key or design-system; brand.md is always available.

Same Visual consistency rule as agents/link.md — derive from the fb.ai brand kit (fivebucks_get_brand_kit, when FIVEBUCKS_API_KEY set) → local design-system/ → brand.md (fallback), never hardcode brand colors/fonts from memory.

Strategic + financial context:

• brands/{brand}/product.md — KPI definitions: what counts as "active user", DAU/WAU/MAU windows, feature adoption metrics, plan tier list
• brands/{brand}/finance.md — KPIs to highlight (MRR, ARR, gross margin, runway, top movers), runway calc method, alert thresholds
• brands/{brand}/investors.md — Investor List, Founder Voice sample paragraphs, Sections to Include, Sections to OMIT, Prior Updates Archive

If brand.md, product.md, finance.md, or investors.md is missing, abort with a failed log and tell the user which file to populate. design-system/ missing is NOT a hard fail — fall back to brand.md and continue.

If investors.md has zero entries in Investor List, abort with failed — there is no audience to send to. Tell the user to populate investors.md first.

Step 2 — Resolve the Reporting Period

Resolve period_start, period_end, period_label:

• Default: previous calendar month. If today = 7 May 2026 → period = 1 Apr 2026 – 30 Apr 2026, label = "April 2026".
• User-supplied: parse to ISO range. Examples: "March 2026" → 1–31 Mar 2026, "Q1 2026" → 1 Jan – 31 Mar 2026.
• Compute prior period for comparison: same length immediately before (e.g., March 2026 for an April update).

Step 3 — Pull Financial Data from Xero

Confirm Xero MCP authentication (the tool requires the connected user's organisation). If not connected, abort and tell the user to connect Xero in Claude settings.

Xero is a deferred MCP tool. Load its schema before calling:

Use ToolSearch:
- query: "select:mcp__claude_ai_Xero__whoami,mcp__claude_ai_Xero__get_connected_user_organisation,mcp__claude_ai_Xero__get_profit_and_loss,mcp__claude_ai_Xero__get_cash_position,mcp__claude_ai_Xero__get_financial_position"

After loading, you can call the Xero tools directly throughout the rest of this skill — they remain in the tool schema for the duration of the run.

Use mcp__claude_ai_Xero__get_connected_user_organisation

Pull P&L for the period:

Use mcp__claude_ai_Xero__get_profit_and_loss:
- from_date: <period_start ISO>
- to_date: <period_end ISO>
- periods: 2  # current period + prior period for comparison

Pull cash position:

Use mcp__claude_ai_Xero__get_cash_position

Compute runway per finance.md Runway Calc Method (e.g., cash / 3-month-trailing average burn). Read the formula from finance.md — never hardcode.

Capture: revenue, gross margin, operating expenses, net loss/profit, cash on hand, runway in months, top expense movers (largest deltas vs prior period).

Step 4 — Pull Stripe MRR / Churn

If the Stripe MCP is not yet authenticated, kick off the OAuth flow first:

Use mcp__claude_ai_Stripe__authenticate

Once authenticated, the Stripe MCP exposes its real tools. Load them via ToolSearch before calling — Stripe tools are deferred until OAuth completes. Pull for the reporting period:

• MRR at period start
• MRR at period end (delta = net new MRR)
• ARR at period end (MRR × 12)
• Active customer count at period end
• New customers added during period
• Churned customers during period (cancellations + downgrades)
• Net revenue retention (NRR) if computable from upgrades/downgrades

If the Stripe MCP cannot return one of these directly, derive from invoice/subscription line items where possible. If a metric truly cannot be computed, mark it "not available this month" — do not fabricate.

Step 5 — Pull Product KPIs from PostHog

Use the active PostHog project (already scoped via the MCP context). Pull DAU/WAU/MAU per the windows defined in brands/{brand}/product.md (e.g., MAU = unique users with event_x in last 30 days).

Use mcp__claude_ai_PostHog__query-run:
- query: <HogQL query computing DAU/WAU/MAU per product.md definition for period_start..period_end>

Compute deltas vs prior period. Pull any other KPIs product.md flags as critical (e.g., "feature adoption %", "trial-to-paid conversion %", "weekly retained user cohort"). Cite the metric definition source in the output.

Step 6 — Pull Customer Wins + Pipeline from Notion CRM

Search the brand's CRM database for closed-won deals + new prospects in the period.

1. Resolve the DB to a data_source_url:

   Use mcp__claude_ai_Notion__notion-fetch:
   - id: "${BRAND}_CRM_DB"
   

   Extract the collection:// URL from the response — typically data_sources[0].url. Save as data_source_url.

   CRM DB bootstrap check — abort cleanly if missing. If ${BRAND}_CRM_DB is not set in .claude/settings.local.json OR the fetch returns not_found / 404, the brand's CRM has not been bootstrapped yet. Do NOT attempt to create it here — that's apollo-lead-prospector's responsibility on its first run. Abort with this user-facing message:

   The brand's CRM database is not yet bootstrapped. Run /link-skills:apollo-lead-prospector first — it creates ${BRAND}_CRM_DB on first execution. Then re-run investor-update-writer.

   Then jump to the Final Step — Log to Dashboard with status: "failed" and a summary of "aborted: ${BRAND}_CRM_DB not bootstrapped — run apollo-lead-prospector first". End the run.

2. Search inside that data source for closed-won deals:

   Use mcp__claude_ai_Notion__notion-search:
   - query: "Stage:Closed Won"
   - data_source_url: <data_source_url from step 1>
   - query_type: "internal"
   

Filter results by close date inside the reporting period. Capture: client name (subject to OMIT rules — see Step 8), tier, ACV, persona, close date.

Also pull new prospects added during the period (Stage = New / Qualified) for pipeline movement context — re-run notion-search with the same data_source_url and an updated query (e.g. "Stage:New" / "Stage:Qualified").

Step 7 — Read Prior Updates Archive (Recurring Themes)

Read the last 3-6 monthly updates from investors.md → Prior Updates Archive section. Identify:

• Outstanding asks — asks raised in prior updates that are still open. Decide whether to repeat (with progress note) or retire (with closure note).
• Recurring wins — themes (e.g., "we keep landing FinTech logos" or "API uptime > 99.95%") that are now baseline rather than newsworthy. Avoid recycling stale wins as fresh.
• Promised-vs-delivered — anything the founder promised in prior updates ("we'll ship Feature X next month") and check whether it shipped. If yes, surface as a delivered win. If no, address candidly in Lowlights.

Step 8 — Compose the Update Sections

Compose update sections per investors.md → "Sections to Include". The standard set is:

Section	Source	Content
TL;DR	All sources synthesized	3 bullets max — biggest win, biggest number, biggest ask. No fluff.
KPIs	Xero + Stripe + PostHog	Period vs prior period table. Include only metrics flagged in finance.md "KPIs to Highlight" + product.md critical KPIs. Show absolute number + delta.
Wins	CRM closed-won + product milestones + press hits	3-5 specific wins. Redact customer names if investors.md "Sections to OMIT" forbids naming customers. Use "a Series-B FinTech in Singapore" style instead.
Lowlights	Honest reflection + alert thresholds from finance.md	2-3 honest items. If runway dropped below finance.md alert threshold, lead with it. If churn spiked, name it. Investors.md mandates honesty.
Asks	Auto-derived + user-supplied custom asks	Specific, actionable asks. Format: "Intros to {company type/role/region}", "hiring help for {role}", "advice on {specific topic}". Pull recurring open asks from Step 7.
Hires	HR data if available, else CRM/Slack signals	Any new joiners (name, role, start date). Departures handled discreetly — name role and timing only, no editorializing.

Match founder voice — read the "Founder Voice" sample paragraphs in investors.md and mirror sentence length, vocabulary, level of self-deprecation, and signature phrases. Do NOT default to corporate-PR voice.

Strip any content that hits "Sections to OMIT" rules. Apply the rules literally — if "specific customer names" is on the OMIT list, redact every customer name in the entire update, including the Wins section.

Total length: aim for 600-900 words. Investor updates that are too long don't get read.

Step 9 — Generate the Branded Deck

Investors increasingly read updates on phones — a branded deck reads better than a wall of email text and signals operational maturity. Gamma deck is the primary deliverable; Google Doc is the fallback when Gamma is unavailable or fails. Mirrors proposal-generator Step 5 and financial-reporter Step 6.

Always save the local markdown audit first — this is the source-of-truth that survives both Gamma and Google Drive failures:

outputs/{brand}/investors/InvestorUpdate_{YYYYMM}.md

Then attempt the deck:

Preferred path — Gamma deck (when Gamma MCP is connected for the brand):

Use mcp__claude_ai_Gamma__generate_from_template:
- input_text: <markdown source from Step 8 — sections become deck cards: TL;DR → KPIs → Wins → Lowlights → Asks → Hires>
- text_options: { "amount": "preserve", "tone": "<from brand.md voice>", "language": "en" }
- card_options: { "dimensions": "16x9" }
- theme_name: "<brand theme if available, else default>"
- additionalInstructions: "Brand: {brand}. Use brand colors: primary {HEX}, accent {HEX}, background {HEX}, text/dark {HEX} (map each from the active source per the Brand kit field map in agents/link.md — fb.ai kit provides primary/accent/dark/background; local design-system/ or brand.md when the kit is null — never invent). Typography: heading {font-family} + body {font-family} (family names from fb.ai kit / local design-system/ when present, brand.md Google Font when fallback; font weight scale comes only from the local design-system, not the kit). Voice: {voice from brand.md}. The deck is a monthly investor update — KPI tables on the KPIs card, bullet wins/lowlights, callout style for the Asks card. Keep visual identity tight to the brand throughout."
- format: "presentation"
- exportAs: "pdf"

Capture the returned gammaUrl (web link) and pdfUrl (PDF export). These become deck_url and pdf_url downstream.

Fallback path — Google Doc (when Gamma MCP is unavailable for the brand, or generate_from_template returns an error):

Use mcp__claude_ai_Google_Drive__create_file:
- name: "InvestorUpdate_{YYYYMM}"
- mimeType: "application/vnd.google-apps.document"
- content: <markdown source from Step 8>
- parents: [<brand's investor folder ID from settings.local.json, or root>]

Capture the returned webViewLink as deck_url (set pdf_url to empty in this case — GDoc readers can export PDF on demand).

On total failure (both Gamma AND Google Drive errored): the local markdown audit at outputs/{brand}/investors/InvestorUpdate_{YYYYMM}.md is still preserved, but investors cannot receive a deliverable URL. Do NOT proceed to Step 10 (Gmail drafts) — sending drafts without a deck link defeats the purpose. Mirror the Step 6 abort pattern: jump directly to the Final Step — Log to Dashboard with status: "failed" and a summary of "aborted: Gamma + Google Drive both unavailable — local markdown audit at outputs/{brand}/investors/InvestorUpdate_{YYYYMM}.md is preserved. Connect Gamma OR Google Drive and re-run.". Skip Steps 10–12 (no Slack notify either — the dashboard log is the failure record). End the run.

Record which path produced the deck for the downstream steps:

• deck_format = "gamma" | "gdoc"
• deck_url = the URL (Gamma share link or Google Doc webViewLink)
• pdf_url = the Gamma PDF export URL (empty when fallback path was used)

Step 10 — Create Gmail Drafts

Pick dispatch mode from inputs (per-investor default, or bcc).

Per-investor mode (default): loop over investors.md → Investor List. For each investor whose preferred update frequency includes this month, create one draft.

Use mcp__claude_ai_Gmail__create_draft:
- to: <investor.email>
- from: <founder email from brand.md>
- subject: "{Brand} Investor Update — {Month YYYY}"
- body: <short personalized opener + TL;DR section + branded deck link from Step 9 + PDF link (when Gamma path) — see structure below>

Email body structure (kept tight — the deck carries the detail):

1. Opening — 1-2 sentences, persona-adapted opener from investors.md per-role salutations (e.g., "Hi {name}, thanks again for the intro to {firm} last month"). Otherwise a generic opener in founder voice.
2. TL;DR — the 3 bullets from Step 8 (biggest win, biggest number, biggest ask) — investors who never open the deck still get the summary in their inbox.
3. Deck link — "Full update with KPI tables, wins, lowlights, and asks: <deck_url from Step 9>". When deck_format = "gamma", also surface the PDF: "Prefer PDF? <pdf_url>".
4. Reply CTA — invite a reply for questions or intro asks.
5. Signature — founder name + title from brand.md.

Do NOT paste the full markdown source into the email — the deck IS the deliverable; the email is the cover note.

Personalize the opening line per investor role if investors.md defines per-role salutations.

BCC mode: create one draft to the founder's own email, with all investors on BCC.

Use mcp__claude_ai_Gmail__create_draft:
- to: <founder email>
- bcc: <comma-separated investor emails from investors.md>
- subject: "{Brand} Investor Update — {Month YYYY}"
- body: <same structure as per-investor mode, generic opener>

Save as draft (do NOT auto-send) unless user explicitly set send_mode = send. The founder reviews and sends manually.

Step 11 — Archive to Notion ${BRAND}_REPORTS_DB

Persist the update as a Notion page in the brand's reports database for permanent archive.

Step 11a — Ensure ${BRAND}_REPORTS_DB exists (first-run only)

This step is primarily for the first-ever investor-update run for a brand. On subsequent runs, the env var is already set and the DB already exists — fetch and proceed.

Read ${BRAND}_REPORTS_DB from .claude/settings.local.json.

IF env var is set:
  fetch the DB → if fetch succeeds → DB exists → DO NOT create. Skip to 11b.
  (only create if fetch returns 404 / not_found, meaning the DB was deleted)

IF env var is NOT set:
  → first-ever run for this brand → create the DB.

Use mcp__claude_ai_Notion__notion-fetch:
- id: "${BRAND}_REPORTS_DB"

If env var is unset (or fetch returns not_found), create:

Use mcp__claude_ai_Notion__notion-create-database:
- parent: { "type": "page_id", "page_id": "<brand_parent_page_id>" }
- title: "{Brand Name} Reports"
- properties: {
    "Name":          { "title": {} },
    "Type":          { "select": { "options": [
                        {"name": "Investor Update"},
                        {"name": "Financial Report"},
                        {"name": "Board Memo"}
                      ] } },
    "Period":        { "rich_text": {} },
    "Status":        { "select": { "options": [
                        {"name": "Draft"},
                        {"name": "Sent"},
                        {"name": "Archived"}
                      ] } },
    "Recipients":    { "number": { "format": "number" } },
    "Created":       { "created_time": {} }
  }

After creation, persist the new DB ID back to .claude/settings.local.json under env.{BRAND}_REPORTS_DB (read existing settings, add the key, preserve all other keys, write back). Notify the user in chat (first-run only).

Step 11b — Create the report page

Use mcp__claude_ai_Notion__notion-create-pages:
- parent: { "database_id": "${BRAND}_REPORTS_DB" }
- pages: [{
    "properties": {
      "Name": "InvestorUpdate_{YYYYMM}",
      "Type": "Investor Update",
      "Period": "{period_label}",
      "Status": "Draft",
      "Recipients": <count from investors.md list>
    },
    "content": "<markdown source from Step 8 + deck URL + deck format (gamma|gdoc) + PDF URL (when gamma) + Gmail draft IDs>"
  }]

Capture the returned page URL.

Step 12 — Notify Slack

DM the founder via Slack so they know drafts are ready to review and send.

Before calling slack_send_message, you MUST first call ToolSearch with query "select:mcp__claude_ai_Slack__slack_send_message" to load the tool schema. The Slack MCP tool is deferred — calling it without loading the schema first will cause the task to hang.

Use mcp__claude_ai_Slack__slack_send_message:
- channel_id: "$SLACK_NOTIFY_USER"
- text: "Investor update drafts ready for {period_label}.
         Deck ({deck_format}): {deck_url}{ ' — PDF: ' + pdf_url if pdf_url else '' }
         Drafts: {N} Gmail drafts pending review.
         Notion archive: {Notion URL}
         Review and send when ready."

Step 13 — Log to Dashboard

See Final Step below.

---

Output format

Save location — local audit:

outputs/{brand}/investors/

Naming convention:

InvestorUpdate_{YYYYMM}.md

Examples:

• InvestorUpdate_202604.md (April 2026 update)
• InvestorUpdate_202603.md (March 2026 update)

Local audit metadata:

---
Date: YYYY-MM-DD
Skill Used: investor-update-writer
Brand: {brand}
Period: {period_label}
Period Start: YYYY-MM-DD
Period End: YYYY-MM-DD
Recipients: <count>
Dispatch Mode: per-investor | bcc
MRR: <number> {currency}
MRR Delta: <signed delta>
Runway Months: <number>
Cash: <number> {currency}
Customer Count: <number>
Churn Count: <number>
Wins Count: <number>
Asks Count: <number>
Deck Format: gamma | gdoc
Deck URL: <url>
PDF URL: <url — empty when deck_format = gdoc>
Notion Report URL: <url>
Gmail Draft IDs: [<ids>]
Status: Draft | Sent | Failed
---

Deliverables produced:

• Markdown source at outputs/{brand}/investors/InvestorUpdate_{YYYYMM}.md (always — survives Gamma/GDrive failures)
• Branded deck — Gamma (primary, returns gammaUrl + pdfUrl) OR Google Doc (fallback, returns webViewLink)
• Gmail drafts (one per investor, or one with BCC) — pending founder review, cover note + TL;DR + deck link
• Notion ${BRAND}_REPORTS_DB archive entry
• Slack notification to $SLACK_NOTIFY_USER

---

Quality checklist

Before finalizing:

• All four required context files read: brand.md + product.md + finance.md + investors.md (design-system/ is the optional fifth read — see separate design-system check below)
• Reporting period resolved correctly (default = prior calendar month; user override respected)
• Every KPI in the update comes from a real data source (Xero / Stripe / PostHog / Notion CRM) — no fabricated numbers
• Runway computed via finance.md Runway Calc Method — not invented
• KPI selection respects finance.md "KPIs to Highlight" + product.md critical KPIs — no random metrics
• Customer names redacted per investors.md "Sections to OMIT" rules — applied consistently across Wins, TL;DR, every section
• No content from investors.md "Sections to OMIT" appears anywhere in the update
• Founder voice mirrors "Founder Voice" sample paragraphs in investors.md — sentence length, vocabulary, signature phrases match
• Lowlights section is honest (2-3 items) — no rosy spin
• Outstanding asks from prior updates (Step 7) reviewed; closed asks marked closed, repeated asks include progress note
• Total update length 600-900 words
• Gmail drafts saved (not auto-sent) unless user explicitly approved sending
• Recipients count matches investors.md Investor List filtered by frequency for this period
• Branded deck produced — Gamma (primary) or Google Doc (fallback); deck_url captured. If both paths failed, agent aborted with failed log and did NOT send drafts.
• Brand visual source resolved in 3-tier order at Step 1: fb.ai brand kit (fivebucks_get_brand_kit, checked first when FIVEBUCKS_API_KEY set) → local brands/{brand}/design-system/ (when present) → brand.md Colors + Google Font; never blocked on a missing key or design-system
• Gamma additionalInstructions carried explicit HEX values + font-family extracted at Step 1 (from fb.ai brand kit / design-system / brand.md) — deck visual identity matches the brand, not Gamma defaults. (N/A when the GDoc fallback path was used.)
• Gmail body is a tight cover note (opener + TL;DR + deck link + reply CTA) — NOT a paste of the full markdown source; the deck is the deliverable
• Notion archive entry includes deck_url, deck_format, and pdf_url (when Gamma)
• Slack notification sent to $SLACK_NOTIFY_USER with the deck URL
• Local audit file written to outputs/{brand}/investors/ — survives both Gamma and Google Drive failures
• Agent run logged to dashboard

---

Final Step — Log to Dashboard

See docs/new_agent_onboarding/metrics-spec.md for the full JSONB contract.

Use gateway MCP tool `fiveagents_log_run`:
- fiveagents_api_key: ${FIVEAGENTS_API_KEY}
- skill: "investor-update-writer"
- brand: "<active-brand>"
- status: "<success|failed>"
- summary: "<1 line, <200 chars>"
- started_at: "<ISO timestamp>"
- completed_at: "<ISO timestamp>"
- metrics: {
    "date": "YYYY-MM-DD",
    "period_label": "<e.g., April 2026>",
    "period_start": "YYYY-MM-DD",
    "period_end": "YYYY-MM-DD",
    "recipients": 0,
    "dispatch_mode": "<per-investor|bcc>",
    "mrr": 0,
    "mrr_delta": 0,
    "arr": 0,
    "runway_months": 0,
    "cash": 0,
    "currency": "<ISO>",
    "customer_count": 0,
    "new_customers": 0,
    "churn_count": 0,
    "dau": 0,
    "wau": 0,
    "mau": 0,
    "wins_count": 0,
    "lowlights_count": 0,
    "asks_count": 0,
    "hires_count": 0,
    "deck_format": "<gamma|gdoc>",
    "deck_url": "<url>",
    "pdf_url": "<url — empty when deck_format = gdoc>",
    "notion_report_url": "<url>",
    "gmail_draft_ids": ["<ids>"],
    "output_path": "outputs/{brand}/investors/",
    "deliverable": "InvestorUpdate_{YYYYMM}.md"
  }