omonuj
UserClaude Code Agent Skills for building, red-teaming and tuning agentic RL evaluation environments — a four-skill pattern (guardian, validation-debugger, score-tuner, iteration-loop) plus a 24-point adversarial reviewer.
Categories
Indexed Skills (53)
fanout-cve-iteration-loop
Orchestrator skill that runs the full push → Oracle → live red-team → rollout → tune cycle for the fanout-cve-rollout task (<TASK_ID>) until Oracle scores 1.0 and the 5 reward-shaped subscores form a dense, correct partial-credit gradient. Delegates to fanout-cve-task-guardian, fanout-cve-validation-debugger, fanout-cve-score-tuner, and horizon-agentic-reviewer. The agent executes every step except triggering the rollout batch.
fanout-cve-score-tuner
Reward-shaping reviewer/tuner for the fanout-cve-rollout task (<TASK_ID>). Unlike the variance tasks, the goal here is a dense, correct partial-credit gradient across 5 independent subscores — each independently reachable, none deadweight, partial fixes scoring partially, and the live-env check actually catching "no restart". Diagnoses gradient problems and picks the smallest safe fix.
fanout-cve-task-guardian
End-to-end maintenance skill for the fanout-cve-rollout task (<TASK_ID>). Owns pre-push safety checks, local quality/anatomy gates, push, Oracle validation, and rollout monitoring. This is a REWARD-SHAPED task (5 independent subscores, 0.2 each) — the goal is a dense, correct partial-credit gradient that guides agents through the full remediation, NOT the avg < <TARGET_MEAN> variance target of the sibling tasks. Calls fanout-cve-validation-debugger when Oracle fails and fanout-cve-score-tuner when the gradient is wrong.
fanout-cve-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the fanout-cve-rollout task (<TASK_ID>). Maps each of the 5 reward-shaped subscores' failures to a root cause and the smallest safe fix. Use after a validation fails before editing anything.
horizon-agentic-reviewer
DISABLED / NULLED for variation-istio-split-brain (user directive). The live red-team step is intentionally skipped — the user runs the equivalent checks themselves and finds an agent-run pass redundant. DO NOT INVOKE this skill or auto-run any step in it. The workflow is: pre-push gates -> push -> hosted Oracle -> (Oracle 1.0) -> tell the user to trigger the eval batch. The content below is retained ONLY as a manual probe reference; it is not part of any loop.
maddy-iteration-loop
Orchestrator skill that runs the full push → Oracle → live red-team → rollout → tune cycle for the maddy-delivery-split-brain task (<TASK_ID>) until Oracle scores 1.0 and rollout avg < <TARGET_MEAN> with both subscores varying. Delegates to maddy-task-guardian, maddy-validation-debugger, maddy-score-tuner, and horizon-agentic-reviewer at the right phases. The agent itself executes every step except triggering the rollout batch.
maddy-score-tuner
Diagnose rollout regressions and tune difficulty for the maddy-delivery-split-brain task (<TASK_ID>). Specializes in keeping avg rollout score < <TARGET_MEAN> with both subscores varying, avoiding deadweight, and not letting the shared drift-controller surface collapse delivery_consistency and observability_governance into one correlated axis. Picks the smallest safe lever for the observed symptom.
maddy-task-guardian
End-to-end maintenance skill for the maddy-delivery-split-brain task (<TASK_ID>). Owns the pre-push safety checks, local quality/anatomy gates, push, Oracle validation handoff, and rollout monitoring. Primary goal is to keep avg rollout score < <TARGET_MEAN> with healthy variance across both subscores. Calls maddy-validation-debugger when Oracle fails and maddy-score-tuner when rollouts regress.
maddy-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the maddy-delivery-split-brain task (<TASK_ID>). Maps observed log patterns to root causes and the smallest safe fix. Use after a validation fails to understand why it failed before editing anything.
oncall-iteration-loop
Orchestrator skill that runs the full push → Oracle → live red-team → rollout → tune cycle for the oncall-schedule-coverage-collapse task (<TASK_ID>) until Oracle scores 1.0 and rollout avg < <TARGET_MEAN> with both subscores varying and decorrelated. Delegates to oncall-task-guardian, oncall-validation-debugger, oncall-score-tuner, and horizon-agentic-reviewer at the right phases. The agent executes every step except triggering the rollout batch.
oncall-score-tuner
Diagnose rollout regressions and tune difficulty for the oncall-schedule-coverage-collapse task (<TASK_ID>). Keeps avg rollout score < <TARGET_MEAN> with both subscores varying and decorrelated (coverage_restored vs monitoring_governance share the drift surface), and never weakens the hidden <metadata-key> / gap-rule variance gates. Picks the smallest safe lever for the observed symptom.
oncall-task-guardian
End-to-end maintenance skill for the oncall-schedule-coverage-collapse task (<TASK_ID>). Owns the pre-push safety checks, local quality/anatomy gates, push, Oracle validation handoff, and rollout monitoring. Primary goal is to keep avg rollout score < <TARGET_MEAN> with healthy, decorrelated variance across both subscores. Calls oncall-validation-debugger when Oracle fails and oncall-score-tuner when rollouts regress.
oncall-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the oncall-schedule-coverage-collapse task (<TASK_ID>). Maps observed log patterns to root causes and the smallest safe fix. Use after a validation fails to understand why it failed before editing anything.
mtls-iteration-loop
Orchestrator skill that runs the full push → Oracle → live red-team → rollout → tune cycle for the variation-fanout-pipeline-break task (<TASK_ID>) until Oracle scores 1.0 and rollout avg < <TARGET_MEAN> with both subscores varying. Delegates to mtls-task-guardian, mtls-validation-debugger, mtls-score-tuner, and horizon-agentic-reviewer at the right phases. The agent itself executes every step except triggering the rollout batch.
mtls-score-tuner
Diagnose rollout regressions and tune difficulty for the variation-fanout-pipeline-break task (<TASK_ID>). Specializes in keeping avg rollout score < <TARGET_MEAN> with both subscores varying, avoiding deadweight, and not introducing hidden coupling between mtls_handshake and trust_governance. Picks the smallest safe lever for the observed symptom.
mtls-task-guardian
End-to-end maintenance skill for the variation-fanout-pipeline-break task (<TASK_ID>). Owns the pre-push safety checks, local quality/anatomy gates, push, Oracle validation handoff, and rollout monitoring. Primary goal is to keep avg rollout score < <TARGET_MEAN> with healthy variance across both subscores. Calls mtls-validation-debugger when Oracle fails and mtls-score-tuner when rollouts regress.
mtls-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the variation-fanout-pipeline-break task (<TASK_ID>). Maps observed log patterns to root causes and the smallest safe fix. Use after a validation fails to understand why it failed before editing anything.
istio-iteration-loop
Orchestrator skill that runs the full push → Oracle → rollout → tune cycle for the variation-istio-split-brain task (<TASK_ID>) until Oracle scores 1.0 and rollout avg < <TARGET_MEAN> with both subscores varying and decorrelated. Delegates to istio-task-guardian, istio-validation-debugger, and istio-score-tuner at the right phases. The live red-team (horizon-agentic-reviewer) is DISABLED by user directive and is never invoked. The agent executes every step except triggering the rollout/eval batch.
istio-score-tuner
Diagnose rollout regressions and tune difficulty for the variation-istio-split-brain task (<TASK_ID>). Specializes in keeping avg rollout score < <TARGET_MEAN> with both subscores varying AND decorrelated, by tightening/loosening the two implemented /etc/bleater gates and the metadata gates — never by stacking new mechanics and never by adding behavioral/traffic checks (no istiod runs). Picks the smallest safe lever for the observed symptom.
istio-task-guardian
End-to-end maintenance skill for the variation-istio-split-brain task (<TASK_ID>). Owns the pre-push safety checks, local quality/anatomy gates, push, Oracle validation handoff, and rollout monitoring. Primary goal is to keep avg rollout score < <TARGET_MEAN> with healthy, decorrelated variance across both subscores. Calls istio-validation-debugger when Oracle fails and istio-score-tuner when rollouts regress.
istio-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the variation-istio-split-brain task (<TASK_ID>). Maps observed log patterns to root causes and the smallest safe fix. Use after a validation fails to understand why it failed before editing anything.
fanout-cve-iteration-loop
Orchestrator skill that runs the full push → Oracle → live red-team → rollout → tune cycle for the fanout-cve-rollout task (<TASK_ID>) until Oracle scores 1.0 and the 5 reward-shaped subscores form a dense, correct partial-credit gradient. Delegates to fanout-cve-task-guardian, fanout-cve-validation-debugger, fanout-cve-score-tuner, and horizon-agentic-reviewer. The agent executes every step except triggering the rollout batch.
fanout-cve-score-tuner
Reward-shaping reviewer/tuner for the fanout-cve-rollout task (<TASK_ID>). Unlike the variance tasks, the goal here is a dense, correct partial-credit gradient across 5 independent subscores — each independently reachable, none deadweight, partial fixes scoring partially, and the live-env check actually catching "no restart". Diagnoses gradient problems and picks the smallest safe fix.
fanout-cve-task-guardian
End-to-end maintenance skill for the fanout-cve-rollout task (<TASK_ID>). Owns pre-push safety checks, local quality/anatomy gates, push, Oracle validation, and rollout monitoring. This is a REWARD-SHAPED task (5 independent subscores, 0.2 each) — the goal is a dense, correct partial-credit gradient that guides agents through the full remediation, NOT the avg < <TARGET_MEAN> variance target of the sibling tasks. Calls fanout-cve-validation-debugger when Oracle fails and fanout-cve-score-tuner when the gradient is wrong.
fanout-cve-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the fanout-cve-rollout task (<TASK_ID>). Maps each of the 5 reward-shaped subscores' failures to a root cause and the smallest safe fix. Use after a validation fails before editing anything.
horizon-agentic-reviewer
DISABLED / NULLED for variation-istio-split-brain (user directive). The live red-team step is intentionally skipped — the user runs the equivalent checks themselves and finds an agent-run pass redundant. DO NOT INVOKE this skill or auto-run any step in it. The workflow is: pre-push gates -> push -> hosted Oracle -> (Oracle 1.0) -> tell the user to trigger the eval batch. The content below is retained ONLY as a manual probe reference; it is not part of any loop.
maddy-iteration-loop
Orchestrator skill that runs the full push → Oracle → live red-team → rollout → tune cycle for the maddy-delivery-split-brain task (<TASK_ID>) until Oracle scores 1.0 and rollout avg < <TARGET_MEAN> with both subscores varying. Delegates to maddy-task-guardian, maddy-validation-debugger, maddy-score-tuner, and horizon-agentic-reviewer at the right phases. The agent itself executes every step except triggering the rollout batch.
maddy-score-tuner
Diagnose rollout regressions and tune difficulty for the maddy-delivery-split-brain task (<TASK_ID>). Specializes in keeping avg rollout score < <TARGET_MEAN> with both subscores varying, avoiding deadweight, and not letting the shared drift-controller surface collapse delivery_consistency and observability_governance into one correlated axis. Picks the smallest safe lever for the observed symptom.
maddy-task-guardian
End-to-end maintenance skill for the maddy-delivery-split-brain task (<TASK_ID>). Owns the pre-push safety checks, local quality/anatomy gates, push, Oracle validation handoff, and rollout monitoring. Primary goal is to keep avg rollout score < <TARGET_MEAN> with healthy variance across both subscores. Calls maddy-validation-debugger when Oracle fails and maddy-score-tuner when rollouts regress.
maddy-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the maddy-delivery-split-brain task (<TASK_ID>). Maps observed log patterns to root causes and the smallest safe fix. Use after a validation fails to understand why it failed before editing anything.
oncall-iteration-loop
Orchestrator skill that runs the full push → Oracle → live red-team → rollout → tune cycle for the oncall-schedule-coverage-collapse task (<TASK_ID>) until Oracle scores 1.0 and rollout avg < <TARGET_MEAN> with both subscores varying and decorrelated. Delegates to oncall-task-guardian, oncall-validation-debugger, oncall-score-tuner, and horizon-agentic-reviewer at the right phases. The agent executes every step except triggering the rollout batch.
oncall-score-tuner
Diagnose rollout regressions and tune difficulty for the oncall-schedule-coverage-collapse task (<TASK_ID>). Keeps avg rollout score < <TARGET_MEAN> with both subscores varying and decorrelated (coverage_restored vs monitoring_governance share the drift surface), and never weakens the hidden <metadata-key> / gap-rule variance gates. Picks the smallest safe lever for the observed symptom.
oncall-task-guardian
End-to-end maintenance skill for the oncall-schedule-coverage-collapse task (<TASK_ID>). Owns the pre-push safety checks, local quality/anatomy gates, push, Oracle validation handoff, and rollout monitoring. Primary goal is to keep avg rollout score < <TARGET_MEAN> with healthy, decorrelated variance across both subscores. Calls oncall-validation-debugger when Oracle fails and oncall-score-tuner when rollouts regress.
oncall-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the oncall-schedule-coverage-collapse task (<TASK_ID>). Maps observed log patterns to root causes and the smallest safe fix. Use after a validation fails to understand why it failed before editing anything.
mtls-iteration-loop
Orchestrator skill that runs the full push → Oracle → live red-team → rollout → tune cycle for the variation-fanout-pipeline-break task (<TASK_ID>) until Oracle scores 1.0 and rollout avg < <TARGET_MEAN> with both subscores varying. Delegates to mtls-task-guardian, mtls-validation-debugger, mtls-score-tuner, and horizon-agentic-reviewer at the right phases. The agent itself executes every step except triggering the rollout batch.
mtls-score-tuner
Diagnose rollout regressions and tune difficulty for the variation-fanout-pipeline-break task (<TASK_ID>). Specializes in keeping avg rollout score < <TARGET_MEAN> with both subscores varying, avoiding deadweight, and not introducing hidden coupling between mtls_handshake and trust_governance. Picks the smallest safe lever for the observed symptom.
mtls-task-guardian
End-to-end maintenance skill for the variation-fanout-pipeline-break task (<TASK_ID>). Owns the pre-push safety checks, local quality/anatomy gates, push, Oracle validation handoff, and rollout monitoring. Primary goal is to keep avg rollout score < <TARGET_MEAN> with healthy variance across both subscores. Calls mtls-validation-debugger when Oracle fails and mtls-score-tuner when rollouts regress.
mtls-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the variation-fanout-pipeline-break task (<TASK_ID>). Maps observed log patterns to root causes and the smallest safe fix. Use after a validation fails to understand why it failed before editing anything.
istio-iteration-loop
Orchestrator skill that runs the full push → Oracle → rollout → tune cycle for the variation-istio-split-brain task (<TASK_ID>) until Oracle scores 1.0 and rollout avg < <TARGET_MEAN> with both subscores varying and decorrelated. Delegates to istio-task-guardian, istio-validation-debugger, and istio-score-tuner at the right phases. The live red-team (horizon-agentic-reviewer) is DISABLED by user directive and is never invoked. The agent executes every step except triggering the rollout/eval batch.
istio-score-tuner
Diagnose rollout regressions and tune difficulty for the variation-istio-split-brain task (<TASK_ID>). Specializes in keeping avg rollout score < <TARGET_MEAN> with both subscores varying AND decorrelated, by tightening/loosening the two implemented /etc/bleater gates and the metadata gates — never by stacking new mechanics and never by adding behavioral/traffic checks (no istiod runs). Picks the smallest safe lever for the observed symptom.
istio-task-guardian
End-to-end maintenance skill for the variation-istio-split-brain task (<TASK_ID>). Owns the pre-push safety checks, local quality/anatomy gates, push, Oracle validation handoff, and rollout monitoring. Primary goal is to keep avg rollout score < <TARGET_MEAN> with healthy, decorrelated variance across both subscores. Calls istio-validation-debugger when Oracle fails and istio-score-tuner when rollouts regress.
istio-validation-debugger
Fetches and interprets validation logs (No-Op and Oracle) for the variation-istio-split-brain task (<TASK_ID>). Maps observed log patterns to root causes and the smallest safe fix. Use after a validation fails to understand why it failed before editing anything.
api-service-scaffold
Scaffold a production-grade backend HTTP service (REST or GraphQL) with the layering, error model, validation, config and health surface already wired. Use when starting a new service or adding a new bounded context to a monorepo. Produces a runnable skeleton that passes lint + a smoke test, not a toy. Language-agnostic playbook with concrete Node/TypeScript and Python/FastAPI recipes.
backend-observability
Instrument a backend service so failures are diagnosable in production — structured logs, RED/USE metrics, and distributed tracing wired through one correlation id. Use when a service has no observability, when incidents take too long to diagnose, or as the observability step of api-service-scaffold. Vendor-neutral (OpenTelemetry), with concrete cardinality and PII guardrails.
building-backends
The foundational guide for how Claude should build backend systems — the rules, defaults, and judgment that apply to every service before reaching for a specific playbook. Read this FIRST for any backend work (new service, new endpoint, data model, auth, background job). Covers architecture, API design, the data layer, error handling, security, concurrency, and testing, and points to the specific skills (api-service-scaffold, db-migration-guardian, backend-observability) for the deep dives.
db-migration-guardian
Plan and apply relational schema migrations safely against a live database, with zero-downtime as the default. Use before writing or running ANY migration that alters a table other services or a running deployment still read/write. Enforces expand-contract, backwards-compatible steps, lock-aware DDL, and a tested rollback. Postgres-first, with MySQL notes.
cicd-pipeline-hardening
Design or harden a CI/CD pipeline so it's fast, reproducible, and can't be turned into a supply-chain attack vector. Use when setting up CI for a repo, when builds are slow/flaky, or when a security review flags the pipeline. Covers stage ordering, caching, least-privilege tokens, pinned actions, artifact provenance, and safe deploy gates. GitHub Actions-first, portable.
container-image-hardening
Build a small, secure, reproducible container image — multi-stage, non-root, pinned, minimal attack surface. Use when writing or fixing a Dockerfile, when images are huge or slow to build, or when a scan flags container CVEs. Covers layer caching order, multi-stage builds, distroless/minimal bases, non-root runtime, and the .dockerignore/secret-leak traps.
iac-terraform-guardian
Write and change infrastructure-as-code (Terraform/OpenTonu) safely — state integrity, plan review, modularity, and no drift or destroy-surprises. Use before authoring or modifying any IaC that manages real cloud resources. Enforces remote locked state, read-the-plan discipline, module boundaries, tagging, and the destroy/replace guardrails that prevent accidental data loss.
building-frontends
The foundational guide for how Claude should build frontend systems — the rules, defaults, and judgment that apply to every UI before reaching for a specific playbook. Read this FIRST for any frontend work (new screen, component, form, data fetching, styling). Covers component architecture, state, data fetching, forms, accessibility, performance, styling, and testing, and points to the specific skills (react-component-builder, frontend-state-architecture, web-performance-audit) for the deep dives.
frontend-state-architecture
Decide where each piece of state lives in a frontend app and stop the "everything in a global store" and "server data in useState" anti-patterns. Use when state is tangled, when adding a feature that shares data across the tree, or when choosing state tooling. Provides a decision tree separating server cache, URL, form, and client UI state, plus data-fetching and normalization rules.
react-component-builder
Build a reusable React component that is accessible, controllable, and correctly typed — not just something that renders. Use when adding a shared/design-system component or refactoring a one-off into something reusable. Enforces the controlled/uncontrolled contract, a11y semantics, forwarded refs, and states beyond the happy path (loading, empty, error, disabled).
web-performance-audit
Diagnose and fix real web performance problems measured against Core Web Vitals, not vibes. Use when a page feels slow, a Lighthouse/CWV score regressed, or before shipping a heavy feature. Walks load-time (LCP/bundle/images), interactivity (INP/main-thread), and layout-stability (CLS) with the specific fix for each, and forbids premature micro-optimization.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.