Standup Coach

The Standup Coach is an opt-in daily assistant that detects when a team member's standup Today section is light on specifics — no issue references, no concrete tasks — and sends them a private Slack DM with personalised task recommendations. Nothing is posted to the team channel.

It is disabled by default. The SM turns it on when the team is ready.

How it works

Trigger: Runs as a step in daily-standup.yml, after the team summary is generated Workflow: daily-standup.yml (coach step)

Daily standup summary generated
  │
  └─ coach.py runs
        │
        ├─ For each team member who submitted a standup:
        │     is_today_vague()?
        │       — Under min_today_words (default: 20)?
        │       — No #NN issue reference?
        │       — No exempt phrase (code review, OOO, ceremony, etc.)?
        │
        ├─ Flagged members only:
        │     Fetch their open sprint issues from GitHub
        │     Read sprint goal from sprint/NNN/meta.md
        │     Read OKRs from workspace/okr/
        │     LLM: "Given these open issues and this sprint goal,
        │            what should Alice focus on today?"
        │
        └─ Private Slack DM to each flagged member
              (optional: DM summary to SM if notify_sm: true)

Detection is deterministic. No LLM is called until a vague standup is found — the check is just text length and issue reference presence.

Recommendations are personalised. The LLM sees only that individual's open sprint issues, the sprint goal, and the team OKRs. It doesn't have visibility into other team members' standups or history.

What "vague" means

A Today section is flagged only when all of these are true:

It is under min_today_words words (default: 20)
It contains no #NN issue reference (when require_issue_ref: true)
It does not contain any of the exempt phrases below

Exempt phrases (never flagged)

Category	Phrases recognised
Reviews	`code review`, `pr review`, `pull request review`
Collaboration	`pairing`, `pair programming`
Ceremonies	`sprint planning`, `sprint review`, `retrospective`, `grooming`, `backlog`
Company events	`all-hands`, `interview`
Leave	`out of office`, `ooo`, `sick`, `sick leave`, `medical`, `holiday`, `travel`
Onboarding	`onboarding`

This list errs on the side of caution — false positives (coaching someone who had a legitimate update) are worse than false negatives (missing a genuinely vague standup).

Example DM

When Alice submits:

## Today
working on things

Scraut detects this is vague (5 words, no issue reference), then fetches her open sprint issues and sends a private DM:

Looks like today could use a bit of focus! Given the sprint goal of shipping
the OAuth integration, #21 (OAuth: Google provider — 8 sp) would be the
most impactful thing to push forward — it's on the critical path for the v1.0
milestone. If that's blocked waiting on anything, #15 (rate limiting — 5 sp)
is self-contained and ready to go. Adding issue numbers in tomorrow's standup
helps the team spot dependencies early — keep it up!

The tone is always supportive. The prompt explicitly forbids the word "vague" and any implication that the person did something wrong.

Configuration

standup_coach:
  enabled: false                # off by default — SM opts in
  min_today_words: 20           # flag Today sections shorter than this
  require_issue_ref: true       # also require at least one #NN reference
  skip_sprint_start_days: 1     # skip coaching on the first N sprint working days
  notify_sm: false              # DM the SM a daily summary of who was coached

See Configuration Reference → standup_coach for full field descriptions.

Privacy model

What	Who sees it
Individual recommendation DM	Only the team member
SM daily summary (if `notify_sm: true`)	Only the SM, as a DM
Public team channel (`slack_channel`)	Nothing — coach never posts here
`.scraut/` files	Nothing written — ephemeral, no stored history

Requirements

Requirement	Why
`SLACK_BOT_TOKEN` GitHub Secret	Required to send DMs. Without it, detection still runs but DMs are silently skipped.
Sprint issues assigned in GitHub	Without assignees on issues, the coach can't personalise recommendations.
`workspace/okr/okr.md` populated	Optional, but improves recommendation quality.
Sprint goal set in `sprint/NNN/meta.md`	Strongly recommended — the single most important input to the recommendation.

When to enable

Good fit for your team if:

Your sprint backlog is well-labelled and issues are assigned to individuals
The team is large enough that the SM can't check in individually every day
You already use the morning DM standup links

Hold off if:

Your backlog is messy or issues are rarely assigned to individuals — recommendations will be too generic to be useful
Your team is small (2–3 people) and the SM already has daily 1:1 context

Start with the sprint goal

Even before enabling the coach, make sure sprint/NNN/meta.md has a clear, specific sprint goal. The LLM uses this as the primary ranking signal for recommendations. "Deliver rate limiting and OAuth to staging" produces far better recommendations than "continue development work".

Scenario: Coach helps a new team member

Characters: Charlie (new developer, joined mid-sprint), Bob (SM)

Situation: Charlie is ramping up and isn't sure which open issues to pick up.

Charlie's standup:

## Today
Ramping up, going to look at the backlog

What happens:

Coach detects: 9 words, no issue reference → flagged
Coach fetches Charlie's assigned open issues: #31 (API error handling, sp:3), #34 (unit tests for auth, sp:2)
Sprint goal: "Ship rate limiting and OAuth to staging"
LLM generates a recommendation ranking #34 first (auth tests needed before OAuth can ship)
Charlie receives a private DM with specific next steps
Bob is not notified unless notify_sm: true

Charlie sees clear guidance without needing a 1:1 with Bob.

How it works​

What "vague" means​

Exempt phrases (never flagged)​

Example DM​

Configuration​

Privacy model​

Requirements​

When to enable​

Scenario: Coach helps a new team member​