Good Morning

Good Morning

Single-command wrapper that runs the user's three morning skills in order. Owns no business logic of its own — every artifact is produced by the delegated skill. This skill exists so the user types one command instead of three.

Quick start

/good-morning                 → run all three phases for default team(s)
/good-morning Aurora          → pass team override down to sprint-* skills
/good-morning --skip-snapshot → skip phase 1 (e.g. snapshot already captured today)

Any extra args (team names, --inc, --sprint, --days, etc.) are forwarded to the sub-skills that recognise them. The wrapper itself only consumes --skip-snapshot, --skip-calendar, --skip-standup, --skip-briefing.

Workflow

- [ ] Phase 1:   sprint-snapshot --phase daily            (Jira → vault canvas + md + jsonl)
- [ ] Phase 1.5: team-calendar OOO digest                 (Confluence ICS → who's out this window)
- [ ] Phase 2:   daily-standup-prep --include-sprint-pulse (per-team standup report + burndown)
- [ ] Phase 3:   daily-briefing --no-focus-blocks         (Outlook 24h recap + today's schedule)
- [ ] Phase 4:   One-line summary of artifacts written

Run phases sequentially, never in parallel. Phase 2 depends on Phase 1's JSONL trend row (burndown forecast reads it). Phase 1.5 runs between them so its OOO digest can flow into both Phase 2 (capacity/coverage) and Phase 3 (FYIs). Phase 3 is independent but ordered last by design.

Auto-continue between phases. When a phase completes, immediately invoke the next one — do not pause for user acknowledgment, do not summarise mid-flight, do not ask "shall I continue?". The user invoked /good-morning expecting the whole chain to run unattended. The only legitimate stop points are:

1. A hard failure in Phase 1 (abort the wrapper).
2. A sub-skill raising a genuinely blocking AskUserQuestion that the wrapper cannot answer on the user's behalf (e.g. missing roster CSV with no source path). In that case, forward the prompt as-is.
3. daily-briefing's stale-carryover prompt or recently-closed re-surface prompt — these are intentional user decisions about specific action items (still open vs done vs cancel vs defer; re-spawn vs false re-surface). Forward as-is; never auto-answer "Still open" on the user's behalf, since silently keeping a stale item is exactly the failure mode these prompts exist to prevent. See daily-briefing/SKILL.md § "Recently-closed dedup rule" and § "Stale carryover rule".

Overwrite prompts, "continue?" prompts, and any other yes/no the wrapper can answer for the user should be auto-answered with the continue/overwrite option — never paused for input.

Phase 1 — Sprint snapshot (daily tag)

Invoke via Skill: sprint-snapshot with --phase daily (preserves literal daily in filenames per sprint-snapshot's phase rule). Forward any team / inc / sprint args the user supplied.

Tell the sub-skill the overwrite behavior up-front so it doesn't prompt: "This run is from /good-morning — if daily.canvas or daily.md already exist for today, overwrite in place. Do not prompt." The JSONL trend log is append-only — sprint-snapshot already handles that correctly.

If sprint-snapshot fails (Jira down, roster missing), surface the error and abort the wrapper — phases 2 and 3 expect the snapshot to exist.

Phase 1.5 — Team-calendar OOO digest

Fetch the team's Confluence Team Calendar (PTO / OOO / Travel / Holiday) and build a short digest of who's out during the briefing window. This is the canonical out-of-office source — without it, vacations are only caught by chance from an email (e.g. John Doe's 6/23–6/29 PTO was missed by the schedule view and only surfaced via a manual status email). The digest flows into both Phase 2 (capacity/coverage) and Phase 3 (FYIs).

Resolve the ICS URL from memory: read reference_team_calendar_ics.md and take the **Team Calendar ICS URL:** line. If the memory is missing, skip this phase with a one-line warning in the Phase 4 summary (⚠ Phase 1.5 skipped: no team-calendar ICS configured — add reference_team_calendar_ics.md) — do not prompt or abort; the wrapper still produces all artifacts without it.

Fetch + parse with a single Bash call (read-only; never write the ICS into the vault). Save the feed to a scratch path under an allowed working dir (C:/temp/teamcal.ics), then parse with Python (PYTHONIOENCODING=utf-8). Parsing rules — these are load-bearing, get them wrong and the digest lies:

• Unfold continuations first — re.sub(r'\n[ \t]', '', text) after normalising \r\n→\n. Confluence wraps long lines.
• All-day events; DTEND is EXCLUSIVE (RFC5545). Subtract one day from DTEND for the human "last day out" — DTSTART 20260629 / DTEND 20260704 = out through 7/3, back 7/4. Getting this wrong reports people back a day early.
• Window filter: keep events where start <= horizon AND end_exclusive > today (Pittsburgh local). Horizon = the current sprint's end_date from _sprint.md if available, else today + 10 days.
• Attribute via the CN= field, NOT SUMMARY. This is the load-bearing parsing fact: Confluence PTO events almost always have a bare SUMMARY:PTO and carry the person's name in the ORGANIZER/ATTENDEE line as CN="Last, First" (e.g. ORGANIZER;X-CONFLUENCE-USER-KEY=...;CN="Doe, John";...). Extract CN="([^"]+)" and match last name AND first name against the team roster CSV ({{rosters_dir}}/<Team>.csv). Matching only SUMMARY finds nobody — that was the bug that let John Doe's PTO go unsurfaced. Fall back to SUMMARY text only for the rare events that do name a person there (Parrott - PTO). Split into attributed (CN or summary matched a roster member) and unattributed (no roster match — counted as "N other platform-org entries", never invented into a roster name).
• Merge adjacent same-person ranges. One person often has several back-to-back PTO VEVENTs (e.g. 6/22–6/26 + 6/29–7/3). Collapse contiguous/overlapping ranges per person so the digest reads "John Doe PTO 6/22 → 7/3" rather than two fragments — otherwise the "last day out" understates the absence.
• Category comes from CATEGORIES (PTO, Holiday, Travel, Conference, Appointment, Work Schedule). Holidays apply to everyone; treat separately from individual PTO.

Build an OOO digest object: for each roster member out in-window → {name, type, first_day, last_day_inclusive} (after merging adjacent ranges per person); plus any team-wide Holiday; plus the unattributed count. Keep it to the roster + holidays — the calendar is the whole platform org (~960 events), so don't dump everything; the digest is just the handful of roster names plus holidays.

Forward the digest to the next two phases (pass it in the invocation text, since sub-skills don't read it themselves yet):

• To Phase 2 (standup): "Team-calendar OOO this window: . Fold into the sprint-pulse capacity read and the scrum-master suggestions — a member who is OOO should not be flagged as 'no activity / confirm blocked', and their in-flight tickets are coverage risks. Reflect reduced effective capacity for OOO days."
• To Phase 3 (briefing): "Team-calendar OOO this window: . Render these as FYI bullets ([[@Person]] OOO {{dates}}) in the briefing, and factor team-wide Holidays into the schedule. Do not spawn action items from OOO (per daily-briefing § 3a — 'reach out to backups' is conditional, not a task)."

If the fetch fails (TLS, network, non-200, empty body), do not abort — emit ⚠ Phase 1.5 skipped: team calendar unreachable (<reason>) in the Phase 4 summary and continue. OOO data is high-value but not load-bearing for the other phases.

Phase 2 — Standup report (with sprint burndown)

Invoke via Skill: daily-standup-prep. Pass IncludeSprintPulse=true explicitly so the clarity-council burndown chart + statistics-expert forecast + scrum-master suggestions run (this is the "sprint burndown" the user wants). Forward team / days args from the wrapper invocation.

Tell the sub-skill: "Overwrite today's standup file(s) without prompting — /good-morning expects idempotent re-runs."

If daily-standup-prep partially fails (e.g. GitLab unreachable), let it produce what it can — its degraded-mode output is still useful. Surface any warnings in Phase 4's summary.

Phase 3 — Daily briefing (report + rollups, no focus blocks)

Invoke via Skill: daily-briefing. Phase 3 must produce all three artifacts:

1. Daily report*— {{vault_root}}\📅\YYYY\MM\YYYY-MM-DDD.md (daily-briefing step 4)
2. Weekly rollup*— {{vault_root}}\📅\YYYY\MM\YYYY-W{NN}.md (daily-briefing step 5)
3. Monthly rollup*— {{vault_root}}\📅\YYYY\YYYY-M{MM}.md (daily-briefing step 6)

Skip step 9 (focus-block proposal) entirely. Tell the sub-skill: "/good-morning is the caller — run steps 1 through 6 (daily report + weekly rollup + monthly rollup). Skip step 9 (focus blocks). Do NOT propose calendar blocks, do NOT create any meeting drafts. Stop after the monthly rollup is written."

Same overwrite rule: if any of the three notes already exist for today, overwrite in place — no prompts.

Phase 4 — Summary

Print one block summarising what was produced:

=== Good morning — 2026-05-18 ===
✓ Snapshot:  [daily.canvas]  [daily.md]              (+1 trend row)
✓ OOO:       John Doe (PTO→6/29), Parrott (PTO→6/23)   (2 out, +1 unattributed)
✓ Standup:   [Aurora 2026-05-D18.md]                 (burndown ↘ on-track)
✓ Briefing:  [2026-05-D18.md]  [2026-W20.md]  [2026-M05.md]  (12 unread, 4 meetings today)

Use markdown links so the user can click straight into the vault. The OOO: line reflects the Phase 1.5 digest — if Phase 1.5 was skipped or the feed was unreachable, replace it with the ⚠ Phase 1.5 skipped: <reason> line instead.

Then run the daily-note section-completeness check (belt-and-suspenders). daily-briefing's § "Pre-save section checklist" should guarantee these, but the chart silently vanished from 4 consecutive June 2026 notes before the checklist existed — so the wrapper re-verifies. Read today's daily note ({{vault_root}}/📅/YYYY/MM/YYYY-MM-DDD.md) and confirm it contains all four mandatory surfaces (see [[feedback_daily_note_mandatory_sections]]):

1. A #### Day at a glance Mermaid gantt block.
2. A #### Email triage at a glance Mermaid pie block.
3. A ### 👥 Team Daily Notes standup wikilink — unless Glob("{{vault_root}}/Scrum Teams/**/YYYY-MM-DD.md") returns zero matches for today.
4. A ### 📊 Cross-day context block with the ![[Daily Notes Dashboard.base#Recent (cards)]] + ![[TODO#By tag]] embeds.

If any is missing, Edit it into the note in place (reconstruct the gantt from the schedule section, the pie from the email-triage counts, the standup link from the Glob result) and append a one-line note to the Phase 4 summary: ↻ Repaired missing daily-note sections: <list>. Do NOT prompt — the sections are mandatory, so repair silently. If this check repairs sections on >1 run, that signals daily-briefing's pre-save checklist is being skipped and needs tightening.

⛔ When reconstructing the gantt, never put a colon (:) inside a task label. Mermaid splits each task line on the first colon to find the :HH-mm, NNmm metadata, so a label like Gap 09:00-10:00 :09-00, 60mm fails with Invalid date. Write time ranges in labels with periods (Gap 09.00-10.00, Lunch gap 12.00-15.00) and keep the only colon in front of the metadata. See daily-briefing/REPORT_TEMPLATE.md § "Day at a glance" for the full rule.

Then run the action-item post-filter pass (belt-and-suspenders). daily-briefing should have applied the verb test and addressed-to-user classifier from its § 3a and the phantom-carryover guard from § 6, but those rules occasionally let an FYI slip through as a - [ ] item. Re-read today's daily note and audit the ## ✅ Open Action Items section:

1. Read {{vault_root}}/📅/YYYY/MM/YYYY-MM-DDD.md and extract every - [ ] line from ### Carrying over (still open) and ### New / from window.
2. For each item, re-apply the heuristics from daily-briefing/SKILL.md § 3a:
   • Verb test — does the bullet name an actionable verb directed at the user (reply / review / approve / merge / sign / decide-with-deadline / schedule / draft / fix / investigate / RSVP)? Status verbs (is, was, has been) fail.
   • Addressed-to-user test — is the ask aimed at the user specifically, or is it broadcast / informational / conditional?
   • OOO / FYI patterns — "reach out to X backups" (X is OOO, no actual task), "read [meeting notes]" with no decision pending, "verify Y was sent" with no time-criticality, "decide: [topic]" with no external deadline or stakeholder waiting.
   • Phantom test — does the bullet contain a person/MR#/Jira ticket that does NOT appear anywhere in today's 📨 Email Triage section or 🗓️ Today's Schedule? If so, there's no fresh signal backing it and it's likely stale.
3. If any items fail the audit, Edit the daily note:
   • Remove the bullet from ## ✅ Open Action Items.
   • Insert it as a plain bullet (no checkbox) under ## 📝 Notes & Follow-ups if the item still has informational value (OOO notice, status awareness).
   • Drop it entirely if it's a phantom with no information content.
4. Append a one-line console note in the Phase 4 summary listing what was reclassified: ↻ Demoted N action items to Notes & Follow-ups: <one-line list with reason per item>. Do NOT prompt — the heuristics are conservative enough to demote silently per user feedback. If the user wants to keep something demoted, they can manually restore it.

This is a sanity net, not the primary classifier. If this step is consistently demoting >2 items per run, that's a signal that daily-briefing/SKILL.md § 3a needs to be tightened further.

Then run a personal-assistant pass. Phase 3 (daily-briefing) was invoked with --no-focus-blocks, so the user never sees a focus-block proposal during /good-morning. To preserve the "what should I actually do next" signal without re-introducing calendar drafts, invoke via Skill agent: with council-single-persona agent, persona pinned to personal-assistant. Pass:

• user_problem: "Given today's standup and briefing artifacts, what are the 3 most important next actions for the user?"
• context: the executive-summary bullets from the daily briefing (read from {{vault_root}}/📅/YYYY/MM/YYYY-MM-DDD.md), the burndown verdict from the standup report, and any impediments/blockers surfaced in Phase 2.
• desired_outcome: "A ## Next actions block — 3 bullets max, each naming a concrete action, timing/deadline, and any dependency or follow-up. No calendar drafts (that's owned by /daily-briefing step 9 when invoked standalone)."
• depth: brief

Append the persona's ## Next actions block to the Phase 4 console output. This makes the wrapper output actionable in its own right rather than just an artifact-link manifest.

Rules

• Sequential, never parallel. Phase 2 reads Phase 1's JSONL; Phase 3 is independent but ordered last by design.
• Auto-continue between handoffs. When one phase finishes, the next starts immediately. No pause, no acknowledgment prompt, no mid-flight summary. Only stop for a hard Phase-1 failure or a genuinely blocking sub-skill question the wrapper can't auto-answer.
• Idempotent. Re-running on the same day overwrites today's artifacts in place. Never prompt — the user has already opted in by invoking /good-morning. The JSONL trend log is the only exception (append-only, by design).
• Phase 3 = three artifacts. Daily report + weekly rollup + monthly rollup. All three. Skip only step 9 (focus blocks).
• No focus blocks, no calendar drafts. Phase 3 is report-only. Do not propose blocks, do not create meeting drafts.
• Forward, don't reinterpret. Team names, --inc, --sprint, --days, etc. are passed through verbatim to whichever sub-skill recognises them. The wrapper doesn't validate them.
• Abort on Phase 1 failure. A missing snapshot breaks Phase 2's burndown. Phases 2 and 3 may degrade gracefully; Phase 1 may not.
• Phase 1.5 is best-effort, never blocking. The team-calendar fetch/parse can fail (TLS, network, ICS schema drift) without affecting the other phases — skip with a ⚠ summary line, never abort. It reads the ICS read-only and never writes the feed into the vault.
• DTEND is exclusive. Whenever the wrapper renders an OOO date range (digest, summary, forwarded text), the "last day out" is DTEND − 1 day. Reporting someone back a day early is the failure mode to avoid.
• No new memory writes. All config (vault root, default team, Jira project, team-calendar ICS URL, etc.) is read from existing memory keys; the wrapper writes no memory files. (The ICS URL lives in reference_team_calendar_ics.md, created once out-of-band — the wrapper only reads it.)

Edge cases

• --skip-snapshot — skip Phase 1 only if the user explicitly opts out (e.g. they already ran it from /sprint-snapshot earlier). Phase 2 will still attempt to read the most recent JSONL row.
• --skip-standup / --skip-briefing — same pattern; document in the Phase 4 summary which phases ran.
• Sub-skill prompts despite the override instruction — answer with the overwrite-equivalent option (Overwrite) on the user's behalf, since /good-morning already implies consent. Do not re-ask.
• Today is a weekend or holiday — still run all three phases; the artifacts are useful for catch-up. daily-briefing's window logic already handles gap days. If Phase 1.5 finds a team-wide Holiday event covering today, note it in the digest so the standup/briefing don't read an empty calendar as "everyone idle".
• Team-calendar ICS missing or unreachable — skip Phase 1.5 with a ⚠ line in the Phase 4 summary; the other phases run unchanged. Add/repair the URL in reference_team_calendar_ics.md.
• --skip-calendar — skip Phase 1.5 explicitly (e.g. the feed is down and the user doesn't want the warning noise). The wrapper consumes this flag alongside --skip-snapshot / --skip-standup / --skip-briefing.
• OOO summary looks wrong by one day — almost always the exclusive-DTEND rule; the last day out is DTEND − 1.