Evaluate Story

Purpose / When to Activate

Activate as the first action of every Delivery Orchestration cycle, before any sub-agent is spawned.

The Orchestrator needs to know:

1. What tier applies (MicroDelivery / Core Team / Core Team + Specialists)
2. Which domains are involved (for specialist registry matching)
3. Whether any pre-flight risk analysis is warranted

Process

1. Read the Story

Read the full Story artefact: acceptance criteria, complexity field, depends_on list, tags.

2. Assess complexity

3. Scan for domain triggers

Read agents/specialists.registry.yaml. Scan the Story's acceptance criteria and title for trigger keywords. Record which specialists would be needed.

4. Verify skill coverage

Read both core/skills/skills-index.yaml and project/skills/skills-index.yaml. For each domain identified in Step 3, verify that at least one skill in either index covers the required capability.

Check:

• Do the identified domains map to existing skills (by description, tags, or category)?
• Are there specialist-triggered domains with no corresponding skill?

If a required skill is missing:

• Record the gap: which capability is needed, for which domain
• This will be reported in the output as skill_gaps

If all required skills exist, record skill_gaps: [].

This step does not block evaluation — it reports gaps. The Orchestrator decides whether to proceed or escalate based on gap severity.

5. Determine tier

complexity ≤ 2  AND  files_affected ≤ 2  AND  criteria_count ≤ 3  AND  no specialist triggers
    → Tier 1: MicroDelivery

complexity 3–7  OR  criteria_count > 3  OR  moderate specialist triggers
    → Tier 2: Core Team

complexity ≥ 8  OR  multiple specialist triggers  OR  cross-cutting at scale
    → Tier 3: Core Team + Specialists

6. Determine pre-flight risk analysis

Risk analysis is warranted when:

• Story touches security, auth, payments, PII
• Schema or API contract changes
• Blast radius is unclear
• Same area has had prior QA failures

Output

Returns (inline, to the Orchestrator's reasoning):

tier: 1 | 2 | 3
specialists_triggered: [list of specialist IDs or empty]
skill_gaps: [list of {domain, missing_capability, severity: critical|non-critical} or empty]
risk_analysis_required: true | false
complexity_assessment: brief note if complexity field is overridden
blocked: true | false  (true if any skill_gap has severity: critical)

This output drives the Orchestrator's next action. It is not written to a file — it is the Orchestrator's internal decision.

Quality Checks

• Tier assignment is consistent with complexity signals
• Specialist triggers are matched against registry, not guessed
• Skill gaps are matched against skills-index.yaml, not guessed
• Over-staffing (Core Team for a typo fix) is as wrong as under-staffing
• A Story with critical skill gaps must not proceed to compose-team