Skip to content

carry-forward-protocol.md

Latest commit

 

History

History
147 lines (100 loc) · 5.37 KB

carry-forward-protocol.md

File metadata and controls

147 lines (100 loc) · 5.37 KB
Raw

Carry-Forward Protocol

How deferred work from Sprint N becomes chains in Sprint N+1. Prevents lost work, forgotten P0s, and perpetual technical debt.

The Problem

Sprints produce carry-forward items: unfinished chains, parked work, P0 discoveries, RED TEAM findings not resolved. Without a protocol, these items get lost between sprints. The ORCHESTRATOR for Sprint N+1 doesn't know what Sprint N left behind.

The Solution

Every sprint produces carry-forward items. Every next sprint consumes them. The sprint registry tracks them across sprints. Nothing falls through the cracks.

Step 1: Extract Carry-Forward (End of Sprint N)

ORCHESTRATOR extracts carry-forward items from three sources:

Source 1: Agent Completion Reports

Read every agent-X-completion.md. Look for:

  • Parked chains — chains the agent started but couldn't finish
  • Blocked chains — chains waiting on another agent's output
  • P0 discoveries — critical issues found during work
  • Suggested follow-up — work the agent recommends for next sprint

Source 2: RED TEAM Findings

Read red-team-findings.md. Look for:

  • Unresolved CRITICAL/HIGH findings — these MUST be carry-forward P1
  • MEDIUM findings not addressed — carry-forward as P2
  • Patterns detected — recurring issues that need structural fixes

Source 3: Sprint Assessment

Read SPRINT-ASSESSMENT.md (your own output). Look for:

  • Failed success criteria — goals not met this sprint
  • Agent FAIL grades — work that wasn't merged
  • Recommendations — structural improvements for next sprint

Step 2: Classify Each Item

For every carry-forward item, assign:

Field Value
Description What the work is (specific, not vague)
Source Which agent/report it came from
Original Priority What priority it had in Sprint N
Carry-Forward Priority What priority it should have in Sprint N+1
Reason Deferred Why it wasn't completed (blocker, scope, time, P0 preempted)
Age How many sprints it has been deferred (1 if first time)

Priority Escalation Rules

Situation Action
P1 item deferred once Stays P1. MUST be in next sprint. Flag to operator.
P1 item deferred twice Escalate to P0. Operator must address before any new work.
P2 item deferred once Stays P2. Should be in next sprint.
P2 item deferred twice Escalate to P1. Must be in next sprint.
P3 item deferred once Stays P3. Include if capacity allows.
P3 item deferred 3+ times Review: is this still relevant? If yes, escalate to P2. If no, remove.
RED TEAM CRITICAL/HIGH Always P1 carry-forward. No exceptions.
P0 discovery not fixed Always P1 carry-forward. Flag to operator.

Step 3: Update the Sprint Registry

Add carry-forward items to the registry at docs/agent-dispatch/sprints/REGISTRY.md:

  1. Update the carry-forward tracker table
  2. Mark items from previous sprints as RESOLVED if completed
  3. Increment the "Age" counter for items still pending
  4. Flag items with age 3+ for operator review

Step 4: Feed Into Next Sprint (Start of Sprint N+1)

When ORCHESTRATOR begins planning Sprint N+1:

Before Phase 1 (Codebase Analysis), read:

  1. sprints/REGISTRY.md — carry-forward tracker
  2. sprint-[N]/SPRINT-ASSESSMENT.md — previous sprint's assessment
  3. sprint-[N]/red-team-findings.md — unresolved findings

During Phase 2 (Work Discovery):

  • All P1 carry-forward items become chains in Sprint N+1 automatically
  • All P2 carry-forward items are candidates (include if scope allows)
  • P3 carry-forward items only if capacity exists after P1/P2 assigned

During Phase 5 (Sprint Proposal):

  • Present carry-forward items separately: "These [N] items are carried from Sprint [N]"
  • Operator can defer again (with documented reason) but cannot silently drop P1 items

Step 5: Document the Bridge

In Sprint N+1's DISPATCH.md, add a section:

## Carry-Forward from Sprint [N]

| Item | Original Sprint | Priority | Assigned To | Chain # |
|------|----------------|----------|-------------|---------|
| [description] | Sprint [N] | P1 | BACKEND | Chain 3 |
| [description] | Sprint [N] | P2 | QA | Chain 5 |
| [description] | Sprint [N-1] → [N] | P1 (escalated) | DATA | Chain 1 |

This makes carry-forward visible. Anyone reading Sprint N+1's dispatch knows exactly what came from previous sprints.

Anti-Patterns

Silent drop: Carry-forward item exists in Sprint N's assessment but doesn't appear in Sprint N+1's dispatch or registry. This means work was lost.

Perpetual defer: Same item deferred sprint after sprint. If an item has been deferred 3+ times, it's either not important (remove it) or being avoided (escalate it).

Priority inflation: Everything carried forward as P1. Defeats the purpose. Only escalate per the rules above.

No registry: Planning Sprint N+1 without reading Sprint N's carry-forward. ORCHESTRATOR MUST read the registry before planning.

Quick Reference

End of Sprint N:
  1. Extract carry-forward from completion reports, RED TEAM findings, assessment
  2. Classify each item (description, source, priority, reason, age)
  3. Update sprint registry

Start of Sprint N+1:
  4. ORCHESTRATOR reads registry FIRST
  5. P1 carry-forward → automatic chains
  6. Document carry-forward in DISPATCH.md