skill-integration-tester

---
name: skill-integration-tester
description: Use when adding or modifying a multi-skill workflow, changing a skill's output format, or verifying that handoffs between chained skills still work before relying on them.
category: Strategy R&D
version: 1.0.0
tools: []
---

# Skill Integration Tester

Validate multi-skill workflows by checking that each named skill exists, that data contracts between consecutive steps are compatible, that file/output naming stays consistent, and that no handoff is broken.

## When to Use

- After adding or modifying a multi-skill workflow
- After changing a skill's output format (fields, structure, naming)
- Before relying on a new workflow, to verify pipeline compatibility
- When debugging why one skill's output isn't working as the next skill's input

## Methodology

1. **Identify the workflow.** Parse the sequence of skills that make up the workflow (e.g. Daily Market Monitoring, Weekly Strategy Review, Earnings Momentum Trading), in order.
2. **Check existence.** Confirm each named skill in the sequence is actually available.
3. **Check contract compatibility.** For each consecutive pair (producer step N, consumer step N+1), verify every field the consumer requires is actually produced by the producer. Mark each handoff PASS, FAIL, or N/A (no data dependency).
4. **Check naming consistency.** Confirm the output naming pattern of step N matches what step N+1 expects to consume.
5. **Dry-run option.** When no real data is available, construct synthetic fixture data shaped like each skill's expected output and validate contract compatibility against that instead — this catches structural mismatches before ever running the real pipeline.

## Diagnosing a Broken Handoff

When a handoff is FAIL, check in order:
1. Does the producer's output actually contain every field the consumer requires?
2. Does the consumer's input accept the producer's output format (types, structure)?
3. Are naming conventions consistent between what the producer emits and what the consumer expects?

## Output

Summarize, per workflow: a step-by-step existence check, PASS/FAIL/N/A for every handoff, any naming-convention violations, and an overall status of valid / broken / warning. Roll up a top-level count across all workflows checked (total, valid, broken, warning) and present it plainly in chat alongside the per-workflow detail.

## Key Principles

1. No credentials required — validation is purely structural
2. Non-destructive — inspect and report only, never modify a skill
3. Deterministic — the same workflow definition always produces the same validation result