The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests production telemetry and metrics (e.g., latency, error rates, and cache hit rates) via the Vercel CLI to identify performance optimization candidates. This constitutes a data ingestion surface where external telemetry influences the generation of optimization recommendations.
Ingestion points: The audit pipeline in TELEMETRY-AUDIT.md and VERCEL-METRICS.md fetches metrics using commands like vercel metrics and vercel usage.
Boundary markers: The skill employs mechanical thresholds and pure-function logic for candidate selection to reduce the risk of adversarial data influencing agent behavior.
Capability inventory: The skill possesses capabilities to execute the Vercel CLI and read repository source files for analysis.
Sanitization: The skill includes a dedicated step in TELEMETRY-AUDIT.md for sanitizing findings and verifying recommendations against the actual source code and observed metrics.
[DYNAMIC_EXECUTION]: The skill uses subprocess execution via the Vercel CLI tool to gather necessary telemetry for its performance audits.
Evidence: VERCEL-METRICS.md documents the use of vercel metrics, vercel whoami, and vercel usage to retrieve project configuration and performance data. This execution is scoped to the skill's primary function of performance auditing.
[EXTERNAL_DOWNLOADS]: The skill references and suggests the integration of several external Node.js packages and development tools.
Evidence: CACHING.md and MEASURE.md recommend libraries such as ioredis, lru-cache, and better-all for managing cache states and asynchronous task concurrency.
Evidence: QUERIES.md provides instructions for pulling and running the official PostgreSQL Docker image (postgres:17) to facilitate local benchmarking and query plan analysis.

performance-and-caching