daily-brief-pattern

name	daily-brief-pattern
description	Shared daily briefing pattern. Defines the orchestration sequence, action selection, and report format for all domain daily briefs (SEO, marketing, CRM, social, content).
metadata	{"version":"1.2.0","git_hash":"c05387b"}

Daily Brief Pattern

Every domain has a daily briefing skill that Ed reads each morning. This pattern defines the shared orchestration sequence, action format, and quality rules. Domain briefs configure it with their own data sources, report paths, and action types.

Principle

The brief is the only thing Ed reads per domain each morning. Everything else in the domain feeds it. Target: Ed reads and executes the actions in 20 minutes over one coffee.

Run Sequence

Every daily brief follows these 8 steps in order:

Step 1: Verify Data Context

Confirm the correct production profile/authentication is active. Domain briefs specify their profile check. Running against a dev/local profile produces silently wrong numbers. Every brief must carry a profile verification line at the top.

Step 2: Check Today's Data Report

Check that the domain's data collection report exists for today. Path pattern: reports/{domain}/daily/{YYYY-MM-DD}/{report-name}.md. If missing, trigger the data collection skill via the Skill tool.

If data collection fails: write a brief that contains the Headline (naming the failure), a System Health row describing the failure, and whichever actions can be generated without the failed data source. Do not halt. Do not ask the user. The master-brief orchestrator depends on every sub-brief returning.

Step 3: Read Strategy Document

Load the domain's strategy master for current phase, priorities, and channel weights.

Step 4: Read Hypotheses

Query the domain's hypothesis ledger:

• in-flight: all active hypotheses
• maturing {today}: hypotheses whose window_end <= today

Step 5: Score Maturing Hypotheses

For each hypothesis where window_end <= today:

1. Pull the current value of metric from today's data report.
2. Compare to baseline + target.
3. Call score {id} {result} {PASS|FAIL|INCONCLUSIVE} {note}.
4. Emit the result in the brief.

Step 6: Generate Actions

Generate all effective actions for today. Minimum 3 when the domain has any in-flight hypotheses or fresh data signal. No maximum. Each action must be backed by a measurable hypothesis AND a data signal that motivates it (a metric trend, an event, a funnel gap). If the domain is genuinely quiet, one maintenance action is acceptable — but prefer to dig until a real signal is found. Selection priority:

1. Dues: actions from previous day's brief that were confirmed but not yet scored.
2. Theme-aligned: 2-3 actions matching the current weekly/phase theme from the strategy master.
3. Channel-balanced: weighted by channel priorities in the strategy master. Do not over-sample one channel unless explicitly prioritised.
4. One experiment: one net-new hypothesis in a channel not tested in >= 7 days. Bias toward learning when data is thin.
5. Floor rules: domain-specific minimum actions when key metrics are zero (e.g., if leads = 0 for 3+ days, force a direct outreach action).

Step 7: Log Actions

Each generated action is logged to the hypothesis ledger via log. Each gets a new {PREFIX}-### ID and a full draft written to data/actions/{PREFIX}-###.md.

Step 8: Write the Brief

Write the brief to reports/{domain}/daily/{YYYY-MM-DD}/{brief-name}.md and print to Ed.

Drop Rules

• No arbitrary maximum on action count. Minimum 3 when the domain has any in-flight hypotheses or fresh data; otherwise at least 1 maintenance action. Actions are hypothesis- and data-driven — if 12 effective actions exist, emit 12.
• No action without a measurable hypothesis.
• No repeat of an approach that failed in the past 14 days (check Dead Hypotheses).
• No action targeting a channel where 3+ hypotheses are already in-flight.
• If the data collection step failed, generate only actions that do not depend on the missing data source, and note the degradation in the Headline. Never return zero actions silently — if no action can be generated, emit exactly one maintenance action (e.g., "investigate {data source} failure") and continue.

Output Format

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

**Profile:** {production profile name}

## Headline
{One sentence: most important thing that changed since yesterday}

## Metric Snapshot

| Metric | Today | 7d ago | 31d ago | Delta (7d) |
|--------|------:|-------:|--------:|-----------:|

## Hypothesis Status

### Maturing Today
| ID | Action | Metric | Baseline | Result | Verdict |
|----|--------|--------|----------|--------|---------|

### In-Flight ({N} total)
{Summary of active hypotheses by channel}

## Today's Actions

### 1. {Action title} [{PREFIX}-###]
**Channel:** {channel}
**Hypothesis:** If {action} then {metric} {direction} within {window}
**Copy-paste ready:**
{The exact text/command/draft Ed needs to execute}

### 2. ...
(repeat for each action — no cap)

Domain Configuration

Config	SEO	Marketing	CRM	Social	Content
Brief name	daily-seo-brief.md	daily-brief.md	daily-crm-brief.md	daily-social-brief.md	daily-content-brief.md
Report path	reports/seo/daily/	reports/marketing/daily/	reports/crm/daily/	reports/social/daily/	reports/content/daily/
Data source skill	daily-seo-brief (self)	lead-tracker	pipeline-tracker	platform APIs	daily-seo-brief (read)
Hypothesis prefix	S-###	H-###	C-###	H-### (shared)	CT-###
Default window	14 days	7 days	14 days	7 days	14 days
Profile check	GSC service account	systemprompt-prod	systemprompt-prod	systemprompt-prod	systemprompt-prod
Floor rule	—	If leads_new_7d = 0 for 3+ days, force direct outreach	If deals_active = 0, force pipeline action	—	If no guide published in 30+ days, force publish action