The Skills Directory
for AI Agents

“Our metric aggregator hit 12,000 calls in the first week. The collections system surfaces our skills alongside complementary observability tools — discovery we never had to build ourselves.”

“We listed a multi-channel notifier and had agents composing it with incident commander within days. Notifications are the missing output layer — every pipeline eventually needs to tell a human something.”

“Our uptime monitor hit 9,000 calls the first week. The incident commander auto-composes with notification and status page skills — agents build reliability pipelines we never designed.”

“Three document processing skills, 17,000 extractions in three weeks. Agents compose our form extractor with DataLens anonymization before downstream processing — a pipeline we never had to market.”

“Our webhook transformer handles 40+ provider formats and agents started composing it with EventMesh routing on day one. The credential vault hit 9,000 calls in two weeks — turns out every pipeline needs secrets management.”

“We listed three code analysis skills and within a week agents were chaining our complexity analyzer with TestHarness mock generation. The collections system put us alongside ComplianceKit audit trails — a pipeline we never imagined but customers love.”

“Our pipeline health checker hit 7,800 calls in the first week. Agents compose it with circuit breaker management and dependency resolution automatically — turns out every production pipeline needs a pre-flight check before every run.”

“Three error analysis skills and 12,000 calls in the first two weeks. Agents compose our error classifier with recovery strategies to auto-remediate failures — the diagnostic layer every pipeline was missing between detection and retry.”

“Our endpoint health prober hit 8,400 calls in the first week. Agents run pre-flight liveness scans before every pipeline execution — turns out nobody wants to start a 12-step workflow only to discover step 3 is down.”

“Our latency benchmark runner hit 7,300 calls in the first week. Agents use it to compare skill alternatives before committing to a pipeline — cold-start detection alone saved one team from a 3x regression they would have shipped blind.”

“Our feature flag evaluator hit 8,900 calls in the first week. Agents use it to gate new skill rollouts and kill-switch unstable integrations — one team went from 45-minute hotfix deploys to instant flag flips for incident response.”

“Our secret rotation scheduler hit 6,200 calls in the first two weeks. Agents compose it with IntegrationBridge credential vaults to auto-rotate before expiry — one team eliminated their monthly credential incident entirely.”

“Our RBAC policy engine hit 7,800 calls in the first week. Agents evaluate access policies in under 10ms before every pipeline action — one team went from ad-hoc permission checks scattered across 12 services to a single policy evaluation point with full audit trails.”

“Our log query engine hit 8,200 calls in the first week. Agents compose it with MetricStream alerting and Observa.ai traces to build full observability stacks — the debugging layer that turns hours of log scrolling into 45ms structured queries.”

“Our data lineage recorder hit 6,700 calls in the first two weeks. Agents compose it with ComplianceKit audit trails and DataLens schema inference to build provenance chains — one team traced a data quality regression across 14 pipeline steps in under a minute.”

“Our adaptive rate limiter hit 9,100 calls in the first week. Agents compose it with PipelineGuard circuit breakers to prevent 429 cascades — one team eliminated their retry storms entirely by adding identity-scoped rate policies at every pipeline ingress point.”

“Our DNS resolution monitor hit 7,800 calls in the first week. Agents compose it with HealthCheck.dev probing and PipelineGuard pre-flight checks to verify network reachability before every pipeline run — one team caught a silent DNS propagation failure that would have broken production for 6 hours.”

“Our input schema validator hit 8,920 calls in the first week. Agents compose it with OutputForge output validation and PipelineGuard pre-flight checks to enforce contracts at every pipeline boundary — one team went from discovering type mismatches at step 7 of a 12-step pipeline to catching them before step 1 executes.”

“Our dependency graph builder hit 6,800 calls in the first week. Agents compose it with CodeAudit.dev risk scanning and PipelineGuard pre-flight checks to audit their own supply chain — one team caught an AGPL dependency three levels deep before it contaminated a commercial pipeline.”

“Our approval gate hit 7,300 calls in the first week. Teams compose it with AccessPolicy.dev RBAC and ComplianceKit audit trails to put a human decision in front of every high-stakes agent action — one enterprise unblocked a deployment that had been stuck in governance review for two quarters.”

“Our code execution sandbox hit 9,800 calls in the first week. Agents compose it with tool-permission-broker grants and egress policies so untrusted code runs with zero standing access — one team promoted agent-written scripts to production for the first time because they could finally prove what the code touched.”

“Our agent identity verifier hit 7,400 calls in the first week. Teams compose it with AccessPolicy.dev RBAC and ApprovalLoop.dev gates so every caller proves who it is before touching a pipeline — one enterprise passed its KYA audit using nothing but the verification records BluePages already stored.”

“Our risk underwriter scored 5,800 invocations in week one. Pipelines compose it with ApprovalLoop gates and SandboxGrid isolation because active guardrails literally lower the premium — one team cut their quoted coverage cost 40% by turning on the guardrails they should have had anyway.”

“Our SLA breach adjudicator ruled on 2,100 disputes in week one. Providers compose it with the refund-policy engine so covered failures resolve instantly and only contested cases escalate — one marketplace cut its dispute backlog to zero because most "disputes" were just refund policies nobody had published.”

“Our eval suite runner hit 5,240 calls in the first week. Teams compose it with the regression gate to block deploys when eval scores drop — one team migrating from OpenAI Evals imported their existing YAML suites directly and had CI-integrated eval gates running the same afternoon.”

“Our CI config generator crossed 5,400 calls in week one — agent pipelines that were hand-maintaining five different YAML dialects now generate them. Teams chain it with the cache optimizer and DeployGuard canaries, so the config that gets generated is also the config that deploys safely.”

“Our residency attestor signed 4,100 region-of-processing attestations in week one. Teams compose it with the region-pinned router so violations are routing decisions instead of audit findings — one enterprise closed a sovereignty procurement blocker by handing the buyer an attestation chain instead of a contract clause.”

“Our run checkpointer took 6,200 calls in the first week — pipelines that used to restart from step one after a crash now resume from the last frontier, and the checkpoint carries the accumulated x402 spend so a recovered run stays inside its original budget. Teams pair it with the crash-resume orchestrator to skip proven steps and only re-run what is unverified.”

“Our usage meter rated 6,800 events in week one and the invoice generator turned a month of x402 receipts into the per-agent, per-pipeline breakdown FinOps kept asking for — the one no provider invoice contains. Teams pair it with the reconciliation auditor to prove metered usage, on-chain settlement, and spending-limit counters are the same number; one caught a double-billed pipeline the day they turned it on.”

“Our cost forecast engine hit 4,820 calls in the first two weeks. Teams compose it with spending limits to set data-driven budget caps instead of guessing — one enterprise found they were over-provisioning three pipelines by 60% and reallocated the savings to new agent workflows.”

“Our workflow starter kit hit 3,840 calls in week one. Teams describe what they want in plain English and get a runnable composition payload with skill selection, step ordering, and cost estimates — one team went from zero pipelines to five production workflows in an afternoon because the templates already wired the error handling and schema mappings they would have spent a week on.”

“Our compatibility checker caught 23 schema mismatches across a six-skill pipeline in under a second — mismatches that would have surfaced as cryptic runtime errors after the team had already wired up payment flows. One platform team now runs integration readiness reports on every new skill before adding it to their approved vendor list — three minutes of pre-flight replaced a week of trial-and-error integration.”

“Our breaking change detector caught a field removal in a version bump that would have broken 14 downstream pipelines — in 85ms, before the release even left CI. One team now gates every skill update on contract compatibility and generates migration adapters for the consumers that need them. They went from uncoordinated surprise breakages to planned, per-consumer upgrade paths with rollback checkpoints.”

“Our deprecation timeline enforcer caught three premature field removals in the first week — fields that were still inside their announced sunset window but had already been stripped from the schema. One team now gates every version bump on deprecation compliance and auto-generates changelogs from schema diffs so their consumers get structured migration notes instead of a Slack message saying "we updated the API."”

“Our version rollout planner crossed 2,300 calls in the first week. Teams compose it with breaking-change-detector and deprecation-timeline-enforcer so a version bump becomes a phased rollout plan instead of a deployment prayer — one team went from "deploy and hope" to consumer-by-consumer migration windows with automatic rollback triggers after their v3 launch broke 11 downstream pipelines.”

“Our OpenAPI linter caught 47 undocumented error responses and 12 missing parameter descriptions in a single specification review — issues that had been silently degrading developer experience for months. One platform team now gates every version bump on spec quality scores and generates visual diffs for PR reviews so schema changes are human-readable instead of raw JSON.”

“Our API reference generator crossed 2,300 calls in week one — teams that were hand-maintaining Markdown docs from OpenAPI specs now generate them with multi-language code examples and cross-linked type definitions. One platform team paired it with the SDK generator to ship a TypeScript client library the same afternoon they published a new API version.”

“Our mock server generator replaced a hand-maintained Postman mock collection that drifted from the spec every other sprint. One team generates WireMock stubs from their OpenAPI spec in CI — their integration tests run against realistic mocks with latency profiles and error injection, no staging environment required. Three services that used to block on each other now develop and test independently.”

“Our integration test generator produced 340 test cases from a single OpenAPI spec in under a second — covering auth flows, pagination edge cases, and error boundaries our manual tests never reached. One platform team paired the flow validator with their deployment pipeline and caught a response schema regression that would have broken three downstream consumers before it reached production.”

“Our synthetic data generator replaced a brittle seed script that took two days to maintain every time the schema changed. One team generates 5,000 referentially-consistent records from their Prisma schema in CI — deterministic seeds mean every test run produces identical data, and the fixture manager catches schema drift before tests start failing with cryptic nulls.”

“Our live response validator caught a silent field deprecation — an upstream team started returning null for a required field after a Friday deploy, and our schema checks flagged it within the hour. The change detector paired with PipelineGuard circuit breakers means our agent pipelines now auto-pause on breaking response changes instead of silently processing garbage data.”