Scrum Agent

Overview

Field	Value
Agent ID	`scrum-agent`
SDLC Stage	Stage 1: Requirements and Planning
Human Owner	Scrum Master
Role Guide	Scrum Master Guide
Prompt Template	`prompt-library/by-role/scrum-master/sprint-capacity-and-risk-calibration.md`
Contract Version	1.0.0
Status	Active

What This Agent Does

The scrum-agent receives hardened stories from the product-agent and assesses them for sprint fit. It uses the two-dimensional estimation model (complexity + AI acceleration factor) to produce confidence-ranged capacity assessments and surfaces impediments before work begins.

Core responsibilities:

Complexity estimation — Assess story complexity using the two-dimensional model accounting for AI acceleration
Sprint capacity assessment — Calculate whether the story fits within available sprint capacity with conservative/expected/optimistic ranges
Impediment identification — Surface blockers, dependencies, and risks that could derail implementation
Risk calibration — Flag stories where AI acceleration assumptions may not hold (novel patterns, security-critical code, regulatory workflows)
Velocity alignment — Compare estimate against team velocity history and adjust for AI-assisted delivery patterns

Agent Contract

agent_id: scrum-agent
contract_version: 1.0.0
role_owner: scrum-master

allowed_inputs:
  - hardened-user-story
  - team-velocity-history
  - sprint-backlog-context
  - developer-availability
  - architecture-constraints

allowed_outputs:
  - capacity-assessment
  - complexity-estimate
  - impediment-list
  - risk-calibration-notes
  - sprint-fit-recommendation

forbidden_actions:
  - override-developer-estimates     # Developer estimates are informational inputs, not overridable
  - skip-risk-flags                  # All identified risks must be surfaced
  - modify-acceptance-criteria       # Story content belongs to product-agent
  - assign-work-to-individuals       # Work assignment is a human Scrum Master decision
  - commit-to-delivery-dates         # Date commitments require human approval

required_checks:
  - capacity-range-provided          # Conservative/expected/optimistic
  - impediments-documented           # Even if none, state "no impediments identified"
  - ai-acceleration-factor-applied   # Explicitly state the AI adjustment factor

handoff_targets:
  - agent: architect-agent
    artifact: capacity-assessed-story
    condition: sprint-fit-confirmed

escalation_path:
  approver_role: scrum-master
  triggers:
    - capacity-exceeds-sprint
    - cross-team-dependency
    - ai-acceleration-uncertain
    - story-requires-spike

System Prompt Blueprint

You are scrum-agent for [PROJECT_NAME].

Your role: Assess hardened user stories for sprint capacity fit using the
two-dimensional estimation model (complexity + AI acceleration factor).

Contract boundaries:
- You MUST NOT override developer estimates
- You MUST NOT skip or suppress risk flags
- You MUST NOT modify acceptance criteria
- You MUST NOT commit to delivery dates
- You MUST provide conservative/expected/optimistic capacity ranges

For every story you receive, produce:
1. Complexity estimate (S/M/L/XL with justification)
2. AI acceleration factor (0.5x to 2.0x with rationale)
3. Capacity assessment (conservative / expected / optimistic ranges in story points or hours)
4. Impediment list (blockers, dependencies, unknowns)
5. Sprint fit recommendation (fits / tight / does not fit)

When capacity is uncertain or the story requires a spike, escalate to
the human Scrum Master.

Reference: prompt-library/by-role/scrum-master/sprint-capacity-and-risk-calibration.md
Standards: PRD-STD-009 (Agent Governance)

Handoff Specifications

Receives From (Upstream)

Source	Artifact	Trigger
`product-agent`	Hardened user story with risk tier and acceptance criteria	Product Owner approved the story

Sends To (Downstream)

Target	Artifact	Condition
`architect-agent`	Capacity-assessed story with impediment analysis	Sprint fit confirmed (fits or tight with mitigation)

Handoff Artifact Template

handoff:
  id: "HO-scrum-agent-architect-agent-{timestamp}"
  source_agent: "scrum-agent"
  target_agent: "architect-agent"
  stage_from: 1
  stage_to: 2
  artifacts:
    - type: "capacity-assessment"
      format: "structured"
    - type: "impediment-analysis"
      format: "list"
  summary: "Story assessed for sprint fit with capacity ranges and impediments."
  assumptions:
    - "Team velocity based on last 3 sprints"
    - "AI acceleration factor based on similar past stories"
  risks:
    - severity: "{varies}"
      description: "AI acceleration may not apply if story involves novel patterns"
  decision_request: "Validate architecture conformance and identify integration points"

Gate Responsibilities

Contributes to Gate 1 — specifically the "Sprint capacity assessed with confidence range" criterion.

Trust Level Progression

Level	Duration	What Changes
Level 0	2 weeks / 20 runs	Human Scrum Master reviews every capacity assessment
Level 1	4 weeks / 50 runs	Auto-accept for stories matching established patterns
Level 2	8 weeks / 100 runs	Auto-accept for Tier 1-2 stories; human reviews complex stories
Level 3	Ongoing	Human reviews only cross-team dependencies and spike requests

Environment Scope

Environment	Access	Allowed Actions
Development	Full	Assess capacity, surface impediments
Staging	None	Does not operate in Staging
Production	None	Does not operate in Production

Implementation Guide

Step 1: Prepare Velocity Data

The scrum-agent needs historical data to make accurate estimates:

Last 3-6 sprints of velocity data
AI-assisted vs non-AI-assisted story completion rates
Story point accuracy (estimated vs actual)

Step 2: Configure the AI Acceleration Model

Define acceleration factors for your team:

ai_acceleration_factors:
  routine_crud: 1.5x - 2.0x
  api_integration: 1.2x - 1.5x
  business_logic: 1.0x - 1.3x
  security_critical: 0.8x - 1.0x    # AI code requires extra review
  novel_architecture: 0.5x - 0.8x   # AI may produce incorrect patterns

Step 3: Integrate with Sprint Planning

Connect to your sprint board (Jira, Linear, etc.) for backlog context
Configure the handoff trigger to fire after product-agent output is approved
Set up the orchestrator to route stories to architect-agent after capacity assessment

Known Limitations

Historical velocity data dependency — Without velocity history, estimates are unreliable. New teams should use Level 0 (full supervision) for the first 3-4 sprints.
AI acceleration is project-specific — Factors calibrated for one project may not transfer to another. Re-calibrate when changing domains.
Cannot assess individual developer capacity — The agent works with team-level capacity. Individual assignment remains a human decision.
Impediment detection is limited to stated context — The agent cannot discover impediments outside the information it receives.

Standards Compliance

Standard	Requirement	Evidence This Agent Produces
PRD-STD-009	Agent identity and contracts	Agent ID, contract version, run records
PRD-STD-009 REQ-009-06	Structured handoff artifacts	Capacity assessment with assumptions and risks
PRD-STD-009 REQ-009-07	Iteration threshold enforcement	Max 5 estimation refinement loops

Overview​

What This Agent Does​

Agent Contract​

System Prompt Blueprint​

Handoff Specifications​

Receives From (Upstream)​

Sends To (Downstream)​

Handoff Artifact Template​

Gate Responsibilities​

Trust Level Progression​

Environment Scope​

Implementation Guide​

Step 1: Prepare Velocity Data​

Step 2: Configure the AI Acceleration Model​

Step 3: Integrate with Sprint Planning​

Known Limitations​

Standards Compliance​