Sprint status narrator

When to run

• Recurring: end of sprint / weekly project checkpoint (recommended)
• "Status on the auth migration" / exec asks "are we on track"

Computer scopes

• Engineering tracker: the project/sprint scope with full item history
• Chat: project channel for blocker context (read)
• Docs: the project brief and stated milestones (the promise being tracked)

Workflow

1. Anchor on the commitment: the sprint goal or milestone doc — status is measured against what was promised, not against motion.
2. Compute the delta since last narrative: completed, in-flight with movement, stalled (no activity beyond threshold), added, removed, re-scoped. Cite item history for each.
3. Detect risk early: stalled critical-path items, review bottlenecks piling on one person, scope growth exceeding burn. State risks as observations with evidence.
4. Pull blocker context from the channel where items reference it; quote the blocker, not vibes.
5. Write the narrative: on-track verdict up top in one sentence, then progress, risks, blockers, scope changes. Plain language; the exec version is the engineer version, shorter.
6. Store the narrative for next run's delta baseline.

Guardrails

• Autonomy: L0. Reads and reports; never edits tracker items, never pings individuals about their velocity. This is a mirror for the team, not surveillance for the org chart.
• No per-person productivity framing — risks are about work items and paths, not people.

Output

Eval fixtures

• Set: 8 sandbox sprint snapshots with seeded ground truth (known stalls, scope adds, a critical-path risk in 3 of them).
• Rubric: delta ledger accuracy; all seeded risks surfaced; zero person-level performance framing; verdict matches seeded reality.
• Ship gate: ledger accuracy ≥ 0.95; risk recall 100% on seeded criticals.