Testing & Agent Automation Skills

Coordinate autonomous workers with bounded locks, deduplication, one-use batons, concurrency permits, shared rate gates, barriers, negative-cache tickets, and renewable freshness leases. Use when parallel agents may duplicate work, exceed a concurrency or API quota, race to consume one action, wait for a phase threshold, or need one temporary refresh owner.

Diagnose covered x402 v2 failures and compare supplied requirements or facilitator capabilities without handling wallet secrets. Use when a buyer or seller integration fails, retry terms may have changed, a facilitator document may not cover the requirement, or the agent needs a normalized cause, known-good recipe, or explicit NOTCOVERED abstention.

Plan and coordinate reliable workflows that combine Agent Enhancer Utilities with another plugin or external tool. Use for repeated, parallel, scheduled, quota-sensitive, freshness-sensitive, duplicate-sensitive, or high-consequence creates, updates, sends, refreshes, imports, and batch jobs where agents must select guards, verify downstream effects, recover from uncertain results, and state the guarantee honestly. Do not invoke for ordinary one-time low-risk reads or writes unless the user explicitly requests a reliability plan.

Compare, check, and temporarily bind normalized x402 requirement snapshots without signing or transferring funds. Use when an agent is about to retry a 402 response, needs to detect changed amount, receiver, network, asset, resource, scheme, or extensions, or must check a supplied facilitator capability document.

Create a short-lived webhook attempt meter, send an authorized test callback to it, and retrieve bounded hash-only delivery evidence once. Use when testing webhook retries, method choice, attempt count, timing, payload stability, or delivery recovery without retaining request bodies, source IPs, or raw headers.

Review bounded MCP tool schemas, names, descriptions, action annotations, initialize capabilities, elicitation modes, and portable JSON Schema edge cases. Use when preparing a connector or plugin for review, diagnosing an MCP capability mismatch, checking whether sensitive fields belong in URL elicitation, or building host-neutral schema tests.

Design and run bounded HTTP failure-path tests with ordered retry sequences, temporary status endpoints, hash-only webhook attempt meters, and deterministic synthetic fixtures. Use when an agent must test retries, Retry-After handling, redirects, webhook delivery attempts, delays, recovery after multiple failures, malformed assumptions, or non-production test data without hosting caller-controlled content.