Daily Master Brief

Daily Master Brief

The one skill Ed runs every morning. Reads all 5 domain briefs and produces a single unified report that answers 9 business questions. This is the operator's command centre — it curates, correlates, and prioritises across all domains.

Principle

Ed should never need to read 5 separate briefs to understand the state of the business. The master brief answers the 9 KPIs in 10 minutes. Domain briefs are drill-down documents — the master brief is the dashboard.

Autonomy Contract (READ BEFORE ACTING)

This skill runs end-to-end without user confirmation between domains. Do not ask Ed "should I run the {X} brief?" — invoke it via the Skill tool and move on.

Exactly two user-facing gates exist in this skill's run:

1. Step 0 — Yesterday Debrief. One structured question batch at the very top. Four grouped prompts in a single invocation. After Ed answers, persist the data and proceed silently.
2. Downstream action execution. When Ed later chooses to execute a proposed action, the downstream skills (email-composer, reddit-reply, content-publish, etc.) run their own approval flows. Not this skill's concern.

Everything in between — data collection, hypothesis scoring, action generation, brief writing — is fully autonomous.

The only permitted unscheduled pause is an auth failure on systemprompt-prod that persists after one retry. Missing data, stale reports, API errors, empty ledgers, failed sub-briefs — all are handled by degrading gracefully and noting the degradation in Section 7 (System Health).

If every sub-brief fails, still produce the master brief. Never return control to Ed without a written report at reports/master/daily/{today}/daily-master-brief.md.

Dependencies (load in order)

1. commons:identity — product positioning, ICP

Also read:

• /var/www/html/systemprompt-web/reports/sales-marketing-strategy.md — cross-domain strategy, funnel definition, domain weights
• All 5 domain briefs for today (see Domain Brief Map below)
• All 4 hypothesis ledgers (SEO, marketing, CRM, content)
• All 5 strategy master documents

CRITICAL: Profile must be systemprompt-prod

Before this skill does anything, verify systemprompt admin session list shows systemprompt-prod as the active profile. Every domain brief enforces this independently, but the master brief must also verify because it reads hypothesis ledger stats directly.

systemprompt admin session switch systemprompt-prod

The 9 KPIs This Brief Must Answer

Every run of this skill must produce a report that explicitly answers:

1. How effective are our strategies? — Hypothesis win rates by domain with trends
2. Are our strategies up to date? — Last-updated timestamps on all strategy masters with staleness warnings
3. How many people have we put in the pipeline? — Pipeline deltas (7d/31d) with new lead counts
4. Where did they come from? — Full funnel attribution: which domain drove which stage
5. How can we improve? — Cross-domain signals: broken funnels, gaps, missed opportunities
6. What is working? — Hypotheses that PASSED today, winning tactics
7. What is not working? — Hypotheses that FAILED today, dead hypotheses
8. What do we need to do today? — All priority actions curated from all domain briefs, ranked by strategic weight, with the top 5 marked PRIORITY
9. Is everything consistent, standardised and updated? — System health: stale briefs, missing reports, overdue hypotheses

If the brief cannot answer any of these, it must say why (missing data, stale brief, failed collection) — never skip silently.

Domain Brief Map

Strategy Master Map

Hypothesis Ledger Map

Note: Social media uses the marketing ledger (H-### prefix). Social H-### rows are a subset of marketing H-###, not additive. When aggregating, do not double-count.

Run Sequence

Step 0: Yesterday Debrief (the only ask gate)

Before any data collection, locate yesterday's master brief at reports/master/daily/{yesterday}/daily-master-brief.md. Extract from Section 6 ("Today's Priority Actions") the full list of action IDs with their one-line titles and domains. If yesterday's brief is missing, fall back to the most recent master brief in reports/master/daily/ and note the gap in the debrief summary.

Then ask Ed a single structured batch of prompts. Use one question-tool invocation — do not drip-feed.

Prompt 1 — Action reconciliation. Present yesterday's proposed actions verbatim:

Yesterday's proposed actions:
  [S-042] Publish tuning guide for Claude Code hooks (SEO)
  [H-118] DM 5 AI governance leads on LinkedIn (Marketing/Social)
  [C-031] Follow up with Acme Corp on evaluation call (CRM)
  [CT-055] Refresh MCP server comparison doc (Content)
  ...

Which did you complete? For each: outcome / result / numbers.
Format: `{ID} done {outcome}` or `{ID} skipped {reason}` or `all done` / `none done`.

Prompt 2 — New data to record. Inbound signals not yet in the system? Replies received, calls taken, new leads, DMs, GitHub stars that became conversations, social engagement numbers, anything else Ed saw yesterday that the trackers wouldn't have caught.

Prompt 3 — External events. Anything happening today that should shape priorities? Launches, deadlines, freezes, planned time-off, customer calls, personal availability.

Prompt 4 — Corrections. Any prior hypothesis, lead, or strategy state Ed wants updated before the briefs run?

Persist before continuing. Parse the answer and write to the correct stores:

• For each confirmed action ID → call done {ID} {outcome} against the matching hypothesis ledger (prefix S- → seo:seo-hypothesis-ledger, H- → commons:marketing-hypothesis-ledger, C- → crm:crm-hypothesis-ledger, CT- → content:content-hypothesis-ledger). For skipped actions → call skip {ID} {reason} on the same ledger.
• For new leads → append via crm:lead-enrichment or crm:pipeline-tracker (match the signal type).
• For external events → write to reports/master/daily/{today}/context.md so Step 7 can read them when ranking priorities.
• For corrections → apply to the named ledger/tracker directly.

Only after the debrief data is persisted does Step 1 proceed. If Ed responds "nothing" / "skip" / "all quiet" to every prompt, persist nothing and continue — but still write Section 0 of the report noting the debrief was skipped.

Step 1: Verify Data Context

Confirm systemprompt-prod profile is active.

Step 2: Check Domain Briefs

For each of the 5 domains, check whether today's brief exists at the expected path.

Trigger order (dependencies flow top-down):

1. seo:daily-seo-brief — must run first (content depends on SEO data)
2. content:daily-content-brief — depends on SEO data
3. commons:daily-marketing-brief — independent
4. social-media:daily-social-brief — reads marketing brief output
5. crm:daily-crm-brief — reads marketing context

For each missing brief:

• Invoke the domain brief via the Skill tool. Do not ask Ed before invoking. Proceed to the next domain when it returns, regardless of whether it succeeded. Failures are logged to Section 7 (System Health), never escalated as questions.
• If invocation fails, fall back to the most recent report for that domain. Mark as STALE: last run {date} in the master brief.
• Never skip a domain entirely. Always show last known state.

Step 3: Read Cross-Domain Strategy

Load sales-marketing-strategy.md for:

• Current domain weights (used to rank actions in Step 7)
• Funnel stage definitions (used for pipeline table in KPI 3)
• Decision framework and priorities

Step 4: Read All Domain Briefs

For each of the 5 domains, extract:

• Headline: the one-sentence summary
• Key metrics: 2-3 numbers with deltas
• Hypothesis status: count of in-flight, count scored today, pass/fail breakdown
• Today's actions: list with hypothesis IDs — every action, not a subset
• Yesterday's execution: actions assigned vs completed (from previous day's brief)

Aggregation contract (MUST). Section 6 of the master brief must contain the union of every action from every domain brief that returned. Never show a single-domain subset. Before moving to Step 5, assert:

1. The master action pool has a contribution from every domain whose brief returned successfully. If a domain returned without actions, that's a finding — add a one-line "Domain {X}: zero actions (reason: {stale data / empty ledger / degrade-mode})" entry to System Health, but the aggregation still counts that domain as represented.
2. If fewer than 5 domains are represented in the pool (counting degrade-mode), treat it as a system-health alert: Section 7 must call out the missing domains at the top.
3. If the pool contains actions from only one domain after all 5 briefs ran, this is a bug — log a master-brief-aggregation-failure warning in the brief's Headline and proceed with what was collected.

Step 5: Read Hypothesis Ledgers Directly

For each of the 4 hypothesis ledgers, run stats to get:

• Total hypotheses logged (all time)
• In-flight count
• Pass/fail/inconclusive counts (all time)
• Win rate percentage
• Trend: compare last 30d win rate to all-time win rate

Step 6: Read Strategy Master Headers

For each of the 5 strategy master documents, read:

• Last updated: YYYY-MM-DD by {skill-name} header
• Current phase
• Compute days since last update
• Flag as STALE if > 7 days since last update

Step 6b: Check GitHub Notifications (run every day)

Fetch unread GitHub notifications using the token from gh auth token (or the github key in .systemprompt/profiles/systemprompt-prod/secrets.json):

GH_TOKEN=$(gh auth token 2>/dev/null)
curl -s -H "Authorization: token ${GH_TOKEN}" \
  -H "Accept: application/vnd.github.v3+json" \
  "https://api.github.com/notifications?all=false&per_page=30"

For each notification, record: [reason] repo | type | title. For any notification on a PR we authored (reason: author), also fetch the PR status (open/closed/merged) and note it.

Add a GitHub Notifications row to Section 7 (System Health) with: total unread count, and a table of each notification with merge/close status for authored PRs. Mark any MERGED PR as a hypothesis partial PASS in the relevant hypothesis notes. Mark any CLOSED-NOT-MERGED PR as a hypothesis partial FAIL.

If gh auth token fails, note "GitHub notifications: auth required" in Section 7 and continue.

Step 7: Collect Key Stats (MANDATORY — never skip)

The Key Stats block is Section 1 of the master brief, immediately after the Yesterday Debrief Summary. These are the numbers Ed glances at every morning before reading anything else. If this section is missing, the brief has failed. Step 11 asserts its presence before the run can complete.

Assemble from today's lead-tracker JSON tail + domain brief headlines.

Sources:

GSC window — source of truth. Two numbers compete: (a) gsc_impressions_7d in funnel-history.jsonl (trailing-partial-week, often stale by 1–3 days) and (b) the aligned-window figure in daily-seo-brief (last complete 7d, bin-aligned per feedback_wow_windows). Always use (b). Near spikes, (a) can invert the true trend. If (a) and (b) disagree, explicitly note the reconciliation in a "Sourcing notes" line under the Content & SEO table.

Known-Null Fields (as of skill v1.3.0). These metrics are currently not captured by lead-tracker and should be emitted as null in metrics.jsonl and labelled not tracked in the Section 1 tables — never 0, never omitted:

• crates_total_lifetime, crates_total_recent_90d (lead-tracker does not hit crates.io download API yet — only referrer uniques)
• If another domain brief fails, any metric sourced from it is null, not 0.

Compute per-metric 7d and 31d deltas by comparing to the same fields in reports/summary/metrics.jsonl (if it has entries from 7 and 31 days ago). If no history exists yet, emit — (baseline) for the delta.

Step 8: Build the Report

Assemble all data into the report structure below, answering each of the 9 KPIs.

Step 9: Write the Master Brief

Write to reports/master/daily/{today}/daily-master-brief.md and print to Ed.

Step 10: Persist Summary Data (MANDATORY — never skip)

Write two outputs to /var/www/html/systemprompt-web/reports/summary/. If either output is missing, the run has failed. Step 11 asserts both files exist before the run can complete.

1. Dated headline snapshot — reports/summary/daily/{YYYY-MM-DD}/headline.md

Contains the full Key Stats block plus per-domain headline numbers. Human-readable, one file per day, never overwritten (if today's file already exists, append a second ## Run at {HH:MM} block — don't clobber).

2. Append-only metrics log — reports/summary/metrics.jsonl

One JSON row per run. Format (schema stable across versions — new fields append, existing fields never rename):

{
  "date": "YYYY-MM-DD",
  "run_at": "YYYY-MM-DDTHH:MM:SSZ",
  "traffic": {
    "website_visitors_7d": 0, "website_visitors_31d": 0,
    "github_views_14d": 0, "github_cloners_14d": 0,
    "github_stars_total": 0, "github_stars_delta_7d": 0,
    "crates_downloads_lifetime": 0, "crates_downloads_recent_90d": 0
  },
  "funnel": {
    "leads_total": 0, "leads_new_7d": 0,
    "deals_active": 0, "pipeline_value_usd": 0,
    "qualified_convos_7d": 0
  },
  "content": {
    "guides_published_total": 0, "guides_indexed": 0,
    "organic_sessions_7d": 0, "gsc_clicks_7d": 0, "gsc_impressions_7d": 0
  },
  "hypotheses": {
    "in_flight_total": 0,
    "win_rate_all_time": 0.0, "win_rate_30d": 0.0,
    "scored_today": 0, "passed_today": 0, "failed_today": 0
  },
  "system_health": {
    "domains_reporting": 0, "domains_stale": 0, "domains_failed": 0
  }
}

If a source is unavailable (e.g., a domain brief failed), emit the field as null rather than omitting it — null means "we tried and couldn't get it today", absent means "this field doesn't exist in this schema version". Both are different from 0.

The jsonl is the long-term trend store. Other skills can read it for sparkline-style reporting without re-scraping historical briefs.

Step 11: Verify Completion Contract (MANDATORY — pre-exit checklist)

Before returning control to Ed, the skill must explicitly assert each of the following. Any failure is a bug and must be reported in the brief's Headline (not buried):

A. The master brief file exists at reports/master/daily/{today}/daily-master-brief.md.

B. Section 1 "Key Stats" is present and populated. Grep the file for ## 1. Key Stats. If absent, the brief has skipped the headline numbers — stop and patch before exit.

C. Section 6 contains the Per-Domain Contribution table. Grep for ### Per-Domain Contribution (or equivalent header). If absent, the aggregation contract has been violated — stop and patch.

D. Section 6 lists at least as many actions as the sum of all domain brief actions (after deduplicating cross-referenced IDs). If only the top 5 appear and the domain contribution table sums to 10+, the brief has truncated. Expand with a compact table of actions #6–N before exit.

E. reports/summary/daily/{today}/headline.md exists and matches the Section 1 numbers.

F. reports/summary/metrics.jsonl has a new row whose date field equals {today} and whose run_at is within the last hour. Do not overwrite — append only.

G. All 9 KPIs have a section or an explicit "Cannot answer: {reason}" line. No KPI may be silently omitted.

H. GSC window reconciled. If the brief quotes a GSC 7d number, it must be the aligned-window figure; any divergence from the funnel-history.jsonl tail must have a one-line reconciliation note.

If any of A–H fails, do not end the run. Patch the brief, re-run the assertions, and only then exit.

Report Structure

# Master Brief -- {YYYY-MM-DD}

**Profile:** systemprompt-prod | **Run at:** {HH:MM}
**Domains reporting:** {N}/5 {list any STALE with last-run date}

---

## 0. Yesterday Debrief Summary

- **Actions confirmed done:** {list of IDs with one-line outcomes}
- **Actions skipped:** {list of IDs with reasons}
- **New data recorded:** {summary of leads/replies/events persisted, with destinations}
- **External events noted:** {summary or "none"}
- **Corrections applied:** {list or "none"}
- **Debrief source:** Ed, {HH:MM today}

If Ed skipped every prompt: "Debrief skipped by operator — no yesterday reconciliation performed."

## 1. Key Stats

> The numbers Ed glances at every morning. Sourced from lead-tracker + domain brief headlines. Deltas computed against `reports/summary/metrics.jsonl`.

### Traffic & Distribution

| Metric | Today | 7d Δ | 31d Δ |
|--------|------:|:----:|:-----:|
| Website sessions (7d rolling) | {N} | {±%} | {±%} |
| Website sessions (31d) | {N} | — | {±%} |
| Unique users (7d) | {N} | {±%} | — |
| Bot share (7d) | {%} | {±pp} | — |
| AI-agent referral sessions (7d) | {N} | {±} | — |
| GitHub views (core+template, 14d dedup) | {N} | {±%} | — |
| GitHub unique cloners (core+template, 14d dedup) | {N} | {±%} | — |
| GitHub stars total (excl self-stars) | {N} | {±} | {±} |
| Crates.io downloads (lifetime, all crates) | {N or "not tracked"} | {±Δ} | — |
| Crates.io downloads (recent 90d, all crates) | {N or "not tracked"} | {±Δ 1d} | — |
| Crates.io referrer uniques (14d) | {N} | {±} | — |

### Funnel

| Metric | Today | 7d Δ | 31d Δ |
|--------|------:|:----:|:-----:|
| Active leads (total) | {N} | {±} | {±} |
| New leads (7d) | {N} | {±%} | {±%} |
| Active deals | {N} | {±} | {±} |
| Pipeline value | ${N} | ±${N} | ±${N} |
| Qualified conversations (7d) | {N} | {±} | {±} |

### Content & SEO

| Metric | Today | 7d Δ | 31d Δ |
|--------|------:|:----:|:-----:|
| Guides published (total) | {N} | {±} | {±} |
| Guides indexed | {N} | {±} | {±} |
| Organic sessions (7d) | {N} | {±%} | {±%} |
| GSC clicks (7d) | {N} | {±%} | {±%} |
| GSC impressions (7d) | {N} | {±%} | {±%} |

If any row is sourced from a failed/stale sub-brief, mark the value `stale:{last-date}` and add the same row to Section 7 (System Health).

## 2. Strategy Effectiveness

> KPI: How effective are our strategies? What is working? What is not working?

| Domain | Prefix | In-Flight | Scored Today | Pass | Fail | Inconclusive | Win Rate (all time) | Trend (30d vs all) |
|--------|--------|----------:|-------------:|-----:|-----:|-------------:|--------------------:|:------------------:|
| SEO | S-### | | | | | | | |
| Marketing | H-### | | | | | | | |
| CRM | C-### | | | | | | | |
| Content | CT-### | | | | | | | |
| **Total** | | | | | | | | |

Note: Social uses H-### (shared with marketing). Social stats are included in the Marketing row.

### What Passed Today

| ID | Domain | Action | Metric | Baseline | Result | Delta | Lesson |
|----|--------|--------|--------|----------|--------|------:|--------|

If zero: "No hypotheses scored today."

### What Failed Today

| ID | Domain | Action | Metric | Baseline | Result | Delta | Lesson |
|----|--------|--------|--------|----------|--------|------:|--------|

If zero: "No failures today."

## 3. Strategy Freshness

> KPI: Are our strategies up to date?

| Document | Last Updated | Updated By | Days Ago | Status |
|----------|:------------|:-----------|--------:|:------:|
| SEO Strategy Master | | | | |
| Marketing Strategy Master | | | | |
| CRM Strategy Master | | | | |
| Content Strategy Master | | | | |
| Sales-Marketing Strategy | | | | |

Status: OK (updated within 7 days) | STALE (7-14 days) | CRITICAL (14+ days)

{If any STALE or CRITICAL: "Action required: update {document} — last touched {N} days ago by {skill}."}

## 4. Pipeline Growth

> KPI: How many people in the pipeline? Where did they come from?

| Stage | Count | 7d Delta | 31d Delta | Primary Source |
|-------|------:|:--------:|:---------:|:---------------|
| Awareness (GSC impressions/wk) | | | | SEO |
| Website Visit (sessions/wk) | | | | SEO + Content |
| GitHub Visit (repo views/14d) | | | | Marketing + Social |
| Clone (unique cloners/14d) | | | | Marketing |
| Active Lead | | | | CRM |
| Qualified Contact | | | | CRM |
| Deal | | | | CRM |

### New Leads This Week

| Name/Company | Source Channel | Date Entered | Current Stage |
|-------------|:--------------|:-------------|:-------------|

Data from CRM brief's pipeline snapshot. If no new leads: "No new leads entered the pipeline this week."

## 5. Cross-Domain Signals

> KPI: How can we improve?

{1-3 specific observations. Each must name concrete numbers and domains. Examples:}

- **Funnel break:** GSC impressions up {N}% but cloners flat — content is attracting visitors who don't convert. Check landing page CTAs.
- **Attribution gap:** {N} new GitHub stars this week but zero referral sessions — README links may be broken.
- **Timing opportunity:** Guide "{slug}" hit page 1 for "{keyword}" — CRM should watch for inbound leads mentioning this topic.
- **Stale channel:** No social actions taken in {N} days despite {N} hypotheses available — social brief may need attention.

If no cross-domain signals: "All domains operating independently. No cross-domain gaps detected."

## 6. Today's Priority Actions

> KPI: What do we need to do today?

**All** actions from all domain briefs, ranked by strategic weight from sales-marketing-strategy.md. No cap — if domains produced 18 hypothesis- and data-driven actions today, show 18. The top 5 are marked PRIORITY so Ed can read top-down and stop when time runs out.

**Per-domain contribution table** (must be emitted before the ranked list):

| Domain | Actions Contributed | Status |
|--------|-------------------:|--------|
| SEO | {N} | {fresh / stale / degrade / failed} |
| Content | {N} | {fresh / stale / degrade / failed} |
| Marketing | {N} | {fresh / stale / degrade / failed} |
| Social | {N} | {fresh / stale / degrade / failed} |
| CRM | {N} | {fresh / stale / degrade / failed} |
| **Total** | **{sum}** | |

If any domain shows 0 actions AND status=fresh, this is anomalous — investigate in Section 7. Single-domain action lists (e.g., all 12 actions from content, zero from the other four) are a bug signature of the aggregation failing; the Headline must flag it.

### 1. [{PREFIX}-###] {title} PRIORITY
**Domain:** {domain} | **Channel:** {channel}
**Why priority:** {one sentence linking to a KPI gap above — strategy effectiveness, pipeline growth, or cross-domain signal}
**Full draft:** See {domain} daily brief, Action {N}

### 2. ... PRIORITY
### 3. ... PRIORITY
### 4. ... PRIORITY
### 5. ... PRIORITY
### 6. ...
### 7. ...
...

**PRIORITY ranking criteria** (applies to the top 5 marker, not a cap on total):
1. Actions that address a cross-domain signal (Section 5) rank highest
2. Actions aligned with the current phase priorities in sales-marketing-strategy.md
3. Actions from domains with declining win rates (need course correction)
4. Dues from yesterday (unfinished business)
5. Experiments in undersampled channels

Actions that don't meet a PRIORITY criterion still appear in rank order below the top 5. Never truncate.

## 7. System Health

> KPI: Is everything consistent, standardised and updated?

| Check | Status | Detail |
|-------|:------:|--------|
| SEO brief ran today | | |
| Marketing brief ran today | | |
| CRM brief ran today | | |
| Social brief ran today | | |
| Content brief ran today | | |
| All strategy docs updated this week | | {list stale ones} |
| Overdue hypotheses (past window_end, not scored) | | {count + list IDs} |
| Orphaned actions (logged but never confirmed) | | {count + list IDs} |
| Data collection errors today | | {list from domain briefs} |
| Hypothesis ledger consistency | | {any gaps in ID sequences} |
| GitHub notifications (unread) | | {count} — {list any authored PRs with merge/close status} |

### Domain Brief Links

- SEO: `reports/seo/daily/{today}/daily-seo-brief.md`
- Marketing: `reports/marketing/daily/{today}/daily-brief.md`
- CRM: `reports/crm/daily/{today}/daily-brief.md`
- Social: `reports/social/daily/{today}/daily-social-brief.md`
- Content: `reports/content/daily/{today}/daily-content-brief.md`

Length target: Under 600 words of prose (tables excluded). This is a 10-minute read. Ed drills into domain briefs for copy-paste-ready drafts and detailed analysis.

Run Modes

daily-master-brief              # Full run: check/trigger all domains, write report
daily-master-brief dry          # Read existing briefs only, do not trigger missing ones
daily-master-brief quick        # Skip Steps 5-6 (hypothesis stats + strategy freshness), faster output

What This Skill Does NOT Do

• Does not generate hypotheses. The master brief curates and prioritises domain-generated actions. It has no ledger of its own.
• Does not score hypotheses. Domain briefs score their own hypotheses in their own run sequences.
• Does not update strategy documents. Strategy masters are updated by their owning domain skills.
• Does not create copy-paste-ready drafts. Those live in domain briefs. The master brief references them.
• Does not replace domain briefs. Each domain brief remains independently runnable. The master brief is a summary layer.

Anti-Sludge Rules

• No vague prose. Every observation must cite a specific number, domain, and metric.
• No motivational language. This is an operator's dashboard, not a pep talk.
• No emojis. Brand voice rules apply.
• Fail loudly. If a domain brief is stale, missing, or errored, the report says so prominently at the top — not buried in a footnote.
• No double-counting. Social H-### hypotheses are a subset of marketing. The total row accounts for this.
• Answer all 9 KPIs or explain why not. If a KPI cannot be answered (e.g., no CRM data), the section must say "Cannot answer: {reason}" rather than being omitted.
• Under 600 words of prose. If longer, it's not a brief — it's a report. Cut the prose, keep the tables.