drusifer/sprint

Full sprint cycle from planning through launch, including all review gates, phase Bloops, and handoff templates.

TLDR: Sprints have three stages: Planning (Cypher→Smith→Morpheus→Mouse), Phase Bloop (*impl per phase), and Close (Oracle→Smith→All retro→Cypher launch). Smith gates after Cypher and Morpheus are mandatory — do not auto-proceed without explicit *user approve. Keep phases small: 1-3 tasks each. Large phases cause context overflow.

Sprint — Full Implementation Cycle

Sprint Stages

Stage 1: Planning      → Cypher → Smith gate → Morpheus → Smith gate → Mouse → Morpheus review
Stage 2: Phase Bloop   → (Neo → Trin → Morpheus) × N phases
Stage 3: Close         → Oracle → Smith → All retro → Cypher launch

---

Stage 1: Sprint Planning

Step 1 — Cypher: Stories + Acceptance Criteria

Cypher *pm plan sprint

• Define user stories with clear acceptance criteria
• Scope the sprint: what's in, what's out
• Hand off to Smith for user review gate

Handoff:

make chat MSG="Stories ready for user review. @Smith *user review <sprint>" PERSONA="Cypher" CMD="pm handoff" TO="Smith"

Gate 1 — Smith: User Story Review

Smith *user review <stories>

• Evaluate stories against HCI principles and user value
• Check: are acceptance criteria testable and user-facing?
• Must post explicit approve or reject

# Approve → proceed to Morpheus
make chat MSG="*user approve. Stories approved. @Morpheus *lead arch sprint" PERSONA="Smith" CMD="user approve" TO="Morpheus"

# Reject → back to Cypher
make chat MSG="*user reject REASON: <reason> | FIX: <fix>. @Cypher revise stories." PERSONA="Smith" CMD="user reject" TO="Cypher"

Step 2 — Morpheus: Architecture

Morpheus *lead arch sprint

• Define architecture decisions and technical design
• Record decisions via @Oracle *ora record decision
• Hand off to Smith for architecture review gate

Handoff:

make chat MSG="Architecture complete. @Smith *user feedback <arch summary>" PERSONA="Morpheus" CMD="lead handoff" TO="Smith"

Gate 2 — Smith: Architecture Review

Smith *user feedback <arch>

• Evaluate architecture for UX impact (flag naming, output formats, breaking changes)
• Must post explicit approve or reject

# Approve → proceed to Mouse
make chat MSG="*user approve. Architecture approved. @Mouse *sm plan sprint" PERSONA="Smith" CMD="user approve" TO="Mouse"

# Reject → back to Morpheus
make chat MSG="*user reject REASON: <reason> | FIX: <fix>. @Morpheus revise arch." PERSONA="Smith" CMD="user reject" TO="Morpheus"

Step 3 — Mouse: Phase Breakdown

Mouse *sm plan sprint

• Break sprint into phases of 1-3 tasks each (no larger — context overflow risk)
• Record phases in agents/mouse.docs/sprint_log.md
• Hand off to Morpheus for plan review

Handoff:

make chat MSG="Sprint planned. Phases ready for review. @Morpheus *lead review sprint plan" PERSONA="Mouse" CMD="sm handoff" TO="Morpheus"

Step 3a — Morpheus: Sprint Plan Review

Morpheus *lead review sprint plan

• Verify phase breakdown aligns with architecture decisions
• Approve or request adjustment to Mouse

Handoff (approved):

make chat MSG="Sprint plan approved. Phase 1 ready. @Neo *swe impl phase-1" PERSONA="Morpheus" CMD="lead handoff" TO="Neo"

---

Stage 2: Phase Bloop

Repeat for each phase until all phases complete.

Step 4 — Neo: Implementation

Neo *swe impl <phase N>

• TDD: write tests first, then implement
• Keep implementation scoped to this phase only

Handoff:

make chat MSG="Phase N impl complete. @Trin *qa uat phase-N" PERSONA="Neo" CMD="swe handoff" TO="Trin"

Step 5 — Trin: UAT

Trin *qa uat <phase N>

• Run tests, verify all acceptance criteria for this phase
• Consult Oracle for expected behavior before asserting

Handoff (pass):

make chat MSG="UAT phase N passed. @Morpheus *lead review phase-N" PERSONA="Trin" CMD="qa handoff" TO="Morpheus"

Handoff (fail):

make chat MSG="UAT phase N FAILED. @Neo *swe fix <issues>" PERSONA="Trin" CMD="qa reject" TO="Neo"

Step 6 — Morpheus: Code Review

Morpheus *lead review <phase N>

• Review for architectural correctness, code quality, maintainability

Handoff (pass — more phases):

make chat MSG="Phase N review passed. @Neo *swe impl phase-N+1" PERSONA="Morpheus" CMD="lead handoff" TO="Neo"

Handoff (pass — last phase):

make chat MSG="All phases reviewed. @Oracle *ora groom" PERSONA="Morpheus" CMD="lead handoff" TO="Oracle"

Handoff (fail):

make chat MSG="Phase N review FAILED. Issues: <issues>. @Neo *swe fix <issues>" PERSONA="Morpheus" CMD="lead reject" TO="Neo"

Fix loop rule: If Neo fails to fix after one retry → Anti-Loop Protocol applies (Oracle consult + user escalation required).

---

Stage 3: Sprint Close

Step 7 — Oracle: Groom

Oracle *ora groom

• Update docs, record decisions, archive sprint artifacts
• Ensure CHAT.md is archived if over 50-100 messages

Handoff:

make chat MSG="Docs groomed. @Smith *user test <sprint>" PERSONA="Oracle" CMD="ora handoff" TO="Smith"

Step 8 — Smith: End-to-End User Testing

Smith *user test <sprint>
Smith *user feedback

• Test all delivered features from the user's perspective
• Apply HCI heuristics — flag rough edges, inconsistencies, confusing interfaces

Handoff (pass):

make chat MSG="User testing passed. @all *sprint retro" PERSONA="Smith" CMD="user approve" TO="all"

Handoff (bug found):

make chat MSG="*user bug CMD: <cmd> | EXPECTED: <x> | ACTUAL: <y> | UX ISSUE: <z>. @Trin triage." PERSONA="Smith" CMD="user bug" TO="Trin"

→ Trin triages → fix loop → re-test before retro

Step 9 — All Personas: Sprint Retrospective

*sprint retro

Each persona posts their domain retrospective to CHAT.md:

| Persona | Retrospective focus | |---------|-------------------| | Neo | Code quality, tech debt, implementation friction | | Trin | Test coverage, regressions caught, test suite health | | Morpheus | Architecture decisions made, anything to revisit | | Oracle | Documentation gaps, decisions not recorded | | Mouse | Phase sizing, blockers, velocity | | Cypher | Story quality, acceptance criteria accuracy | | Smith | UX issues, HCI gaps, user feedback themes |

Output feeds Cypher's backlog before launch.

# Each persona posts:
make chat MSG="<persona> retro: <findings>. Backlog items: <items>" PERSONA="<Name>" CMD="retro" TO="Cypher"

Step 10 — Cypher: Launch

Cypher *pm launch <sprint>

• Announce release
• Add retro feedback to backlog
• Update changelog
• Close sprint

Handoff:

make chat MSG="*pm launch <sprint>. Sprint complete." PERSONA="Cypher" CMD="pm launch" TO="all"

Sprint is complete when Cypher posts *pm launch.

---

Quick Reference

| Step | Persona | Command | Gate | |------|---------|---------|------| | 1 | Cypher | *pm plan sprint | → Smith *user review | | 1a | Smith | *user approve / *user reject | Must approve to proceed | | 2 | Morpheus | *lead arch sprint | → Smith *user feedback | | 2a | Smith | *user approve / *user reject | Must approve to proceed | | 3 | Mouse | *sm plan sprint | → Morpheus *lead review sprint plan | | 3a | Morpheus | *lead review sprint plan | Approve to start phase Bloop | | 4 | Neo | *swe impl <phase N> | → Trin UAT | | 5 | Trin | *qa uat <phase N> | → Morpheus review | | 6 | Morpheus | *lead review <phase N> | Pass → next phase or Oracle | | 7 | Oracle | *ora groom | → Smith testing | | 8 | Smith | *user test + *user feedback | Pass → retro; bug → fix Bloop | | 9 | All | *sprint retro | Feed backlog to Cypher | | 10 | Cypher | *pm launch <sprint> | Sprint complete |

---

Rules

• Short phases: 1-3 tasks each. Large phases cause context overflow.
• No skipping gates: Smith's gates after steps 1 and 2 are mandatory. Never auto-proceed.
• Fix Bloop scope: Fix Bloop targets the failing phase only — never restart the full sprint.
• State saves: Every persona saves state before every handoff (see bob-protocol State Management).
• Chat first: Post the handoff make chat call BEFORE switching. The next persona reads CHAT.md on entry.
• Retro is required: Step 9 is not optional. Retro output is the input to the next sprint's backlog.