The Agent Skills Directory

[SAFE]: The skill is primarily instructional, providing detailed workflows and playbooks for performance engineering across various domains (frontend, backend, database, cloud). It emphasizes evidence-based decision making and measurement over arbitrary code modification.
[NO_CODE]: The provided Python script (scripts/validate_scenarios.py) is a local utility used to ensure the skill documentation contains required sections and keyword coverage. It performs static analysis of local files without executing untrusted code or making network requests.
[INDIRECT_PROMPT_INJECTION]: As an optimization tool, the skill naturally consumes external artifacts such as performance traces, query plans, and metrics. While this represents a potential attack surface for indirect prompt injection, the skill mitigates this through a strong evidence-based workflow that requires human verification and reproducible measurements.
Ingestion points: Processes external metrics, traces, and database query plans (SKILL.md).
Boundary markers: No explicit delimiter instructions for these data types, though the manual verification workflow provides a natural guardrail.
Capability inventory: The skill suggests running analytical commands such as EXPLAIN ANALYZE (references/database-and-cache.md) but does not include automated execution of these commands.
Sanitization: The instructions emphasize using actual production artifacts for evidence-based conclusions rather than relying on unverified inputs.

performance-optimization