Add agent memory retrieval safety skill

[milos-plavsic]

What does this PR do?

Adds a focused Agent Skill for reviewing CockroachDB-backed semantic-memory retrieval before an application treats ranked memories as trustworthy.

The workflow separates four commonly conflated control planes:

• application eligibility predicates
• database grants and roles
• row-level security, when used
• relational integrity constraints

It also covers embedding-space compatibility, bounded retrieval, non-executing plan inspection, synthetic boundary tests, explicit abstention, and a redacted report template.

This is a draft pending maintainer alignment on the proposal and scope.

Related Issue

Closes #19

Type of Change

• New skill
• Skill update/improvement
• Documentation update
• Infrastructure/tooling change
• Bug fix

Skill Compliance Checklist

• SKILL.md includes required frontmatter fields (name, description)
• Skill name follows naming conventions (lowercase, hyphens, gerund form preferred)
• Skill name matches directory name exactly
• Description is specific and includes "when to use" trigger keywords
• Description is max 1024 characters
• Skill is under 500 lines (or uses references/ for detailed content)
• References official CockroachDB documentation (not duplicate content)
• Includes safety guardrails for risky operations (if applicable)
• Tested with at least one AI agent (manual validation)
• No time-sensitive information (version numbers, dates, "currently")
• Directory structure follows specification (only scripts/, references/, assets/ subdirs)
• No reserved words in skill name ("anthropic", "claude")
• Local validation passes: python scripts/validate-spec.py skills/

Documentation Updates

• README.md
• CONTRIBUTING.md
• Skill SKILL.md files
• Other: reusable references/report-template.md

Testing

Manual testing:

• Tested with AI agent: OpenAI Codex, using a multi-tenant vector-retrieval review scenario
• Verified skill is discoverable with prompts about cross-tenant semantic-memory retrieval and embedding migrations
• Validated technical accuracy against CockroachDB vector, EXPLAIN, authorization, and row-level-security documentation
• Tested commands: confirmed the example uses non-executing EXPLAIN; the skill intentionally provides no mutation script

Automated testing:

• Agent Skills quick_validate.py passes
• Local repository validation passes: python scripts/validate-spec.py skills/cockroachdb-application-development/verifying-agent-memory-retrieval-safety/
• CI validation will run automatically on this PR

The repository validator emits a non-blocking suggestion because its gerund allow-list does not recognize verifying, although the name is already a gerund and follows the proposal in #19.

Reviewer Notes

Please focus on whether the boundary between application retrieval policy and database authorization is crisp enough, and whether the read-only EXPLAIN guidance is appropriately conservative.

---

By submitting this PR, I confirm:

• I have read the contributing guidelines
• I have followed the Agent Skills Specification
• I have tested my changes
• I am willing to address review feedback