06-workshops - [Bug] 07-Guardrails in Policy: SensitiveInformation Cedar syntax rejected by API — docs contradict runtime validation

**In which component is this bug present?**

- [ ] 01-AgentCore-runtime
- [x] 02-AgentCore-gateway
- [ ] 03-AgentCore-identity
- [ ] 04-AgentCore-memory
- [ ] 05-AgentCore-tools
- [ ] 06-AgentCore-observability
- [ ] 07-AgentCore-E2E

**Bug Description**

File: `deploy.py`, lines 509–533

Symptom:
```
botocore.errorfactory.ValidationException: An error occurred (ValidationException)
when calling the CreatePolicy operation: SensitiveInformation guardrail requires an
aggregation method call (e.g. count(), categories(), maxConfidenceScore(), minConfidenceScore()).
Per-category indexing is not supported for SensitiveInformation.
```

Root Cause: The deploy script uses per-category indexing syntax for SensitiveInformation policies. All attempted syntax variations are rejected:
- `["US_SOCIAL_SECURITY_NUMBER"].confidenceScore.greaterThanOrEqual(...)` — Rejected
- `.maxConfidenceScore().greaterThanOrEqual(decimal("0.5"))` — Rejected
- `.count() > 0` — Rejected
- `.categories().contains("US_SOCIAL_SECURITY_NUMBER")` — Rejected

**Documentation Conflict:** The official documentation shows an example using per-category indexing for SensitiveInformation, which contradicts the API's runtime validation. Either the docs are wrong or the API has a bug.

Impact: The `block_ssn` and `block_credit_cards` guardrail policies cannot be created. ContentFilter (VIOLENCE) and PromptAttack (JAILBREAK) work correctly.

**Suggested Fix**
Determine the correct Cedar syntax for SensitiveInformation guardrails and update both `deploy.py` and the documentation. If SensitiveInformation is not yet supported via the CreatePolicy API, document that limitation.