Claude Skills by Miosa-osa

> Define a new specialist agent with role, capabilities, triggers, and routing rules.

> Define a new operation (a packaged workflow with agents, skills, and deliverables).

> List all defined operations with their status, agents, and deliverables.

> Validate agent definitions, skill files, and operation configs for correctness.

Evaluate retrieval and generation quality in RAG pipelines. Separate scoring for retrieval (recall, precision, MRR) and generation (faithfulness, relevance, completeness). End-to-end pipeline assessment with bottleneck identification. Triggers on: "eval rag", "rag evaluation", "retrieval evaluation", "rag quality", "rag metrics"

Design binary pass/fail LLM-as-Judge evaluators. Structured prompt engineering for evaluation: criteria definition, rubric construction, few-shot calibration, and bias mitigation. Produces a ready-to-deploy judge prompt with scoring instructions. Triggers on: "judge prompt", "llm judge", "evaluator prompt", "scoring prompt", "grading rubric"

Language Agent Tree Search - Monte Carlo planning - 92.7% on HumanEval

Self-learning system based on SICA, VIGIL, and Mem0 patterns. Auto-triggers after task completion. Captures patterns, consolidates memory, generates skills, recovers from errors.

Self-improving prompts through meta-level optimization

Optimize token usage through prompt caching and compression

Thought-Action-Observation loop for transparent reasoning

Self-correction via critique loop - 18.5 percentage point improvement

Sample multiple paths, select most consistent - +17.9% on GSM8K

Parallel generation through skeleton-first approach for 2x speedup

Generate diverse synthetic test inputs via dimension-based tuple generation. Defines variation dimensions, enumerates combinations, filters for relevance, and produces labeled test cases. For LLM eval pipelines, training data augmentation, and stress testing. Triggers on: "synthetic data", "generate test data", "test inputs", "data generation", "augment data"

Multi-path reasoning with evaluation and backtracking - 74% success on complex tasks

Calibrate LLM-as-Judge evaluators against human labels. Computes TPR, TNR, precision, recall, F1, and Cohen's kappa. Detects systematic biases and recommends prompt corrections. Produces a calibration report with confidence intervals. Triggers on: "validate evaluator", "calibrate judge", "judge accuracy", "evaluator validation", "judge metrics"

Multi-domain audit with weighted scoring. Spawns parallel subagents per audit domain. Each check has severity weight and category weight. Produces a quantified health score (0-100) with prioritized findings. Supports security, code quality, performance, compliance, and custom domains. Triggers on: "audit", "assess", "evaluate quality", "score"

Guided analysis of LLM traces to categorize and pattern-match failures. Reads execution traces, classifies failure modes, detects recurring patterns, and produces actionable taxonomy of errors. For diagnosing why an LLM system fails and where to focus improvement effort. Triggers on: "error analysis", "analyze failures", "failure patterns", "trace analysis", "debug eval"

Audit an LLM evaluation pipeline for correctness, coverage, and reliability. 6 diagnostic areas with structured Check/Finding output. Produces prioritized findings by severity and recommends next skills to run. Catches common eval pitfalls before they corrupt your metrics. Triggers on: "eval audit", "audit evals", "evaluation audit", "check eval pipeline", "eval health"

9 graph analysis operations for knowledge networks: triangles (synthesis opportunities), bridges (critical connectors), clusters (isolated subgraphs), hubs (high-degree nodes), siblings (unconnected items sharing topics), forward/backward traversal, orphans, and staleness scan. The analytical lens on your knowledge structure. Triggers on: "graph", "connections", "network analysis", "knowledge map"

Workspace health diagnostics. Runs targeted checks against the knowledge base: orphaned content, stale signals, missing cross-references, index drift, duplicate detection, broken references, embedding coverage, and quality distribution. Color-coded severity output. Triggers on: "health", "diagnose", "check health", "knowledge base status"

Workspace metrics dashboard. Tracks growth rate, connection density, pipeline throughput, health score trends, budget consumption, and agent utilization over time. Includes trend detection for spotting degradation or acceleration. Triggers on: "stats", "metrics", "dashboard", "workspace numbers"

> Revise existing content for clarity, accuracy, and signal-to-noise ratio.

Create animation-rich HTML presentations from scratch or convert PowerPoint. Progressive disclosure pipeline: content discovery, style discovery, generation. Zero-dependency single HTML file output. Anti-AI-slop guardrails for natural, professional presentations. Triggers on: "slides", "presentation", "slide deck", "create slides", "convert pptx", "html presentation"

> Compress content to a target length while preserving key signal.

> Convert content between genres while preserving the core signal.

> Generate content in the correct genre for the target receiver.

Visual task management dashboard. Terminal-based kanban board, tiled agent status view, and progress tracking. Shows all agents, their current tasks, budget consumption, and overall workspace health at a glance. Triggers on: "board", "dashboard", "kanban", "status board", "show agents"

Atomic task locking. Ensures only one agent works on a task at a time. Returns 409 Conflict if already locked. Auto-releases on agent death or timeout. Prevents double-work and wasted compute in multi-agent systems. Triggers on: "checkout", "lock task", "claim task", "reserve"

Assign work to a specific workspace agent or external runtime. Picks the right adapter (Claude for reasoning, Codex for bulk changes, Gemini for multimodal). Creates tasks with parent chain tracking for full delegation lineage. Triggers on: "delegate", "assign", "hand off", "send to agent"

Agent wake-up and health monitoring protocol. 9-step startup cycle that grounds the agent in identity, fetches tasks, selects work, and begins execution. Also serves as a periodic health check — detect stalled or dead agents. Scheduled or event-triggered. Triggers on: "heartbeat", "wake up", "agent health", "check agents"

Point-to-point messaging between agents. Send, broadcast, receive, and peek at messages. File-based implementation using JSON in inbox directories with atomic writes. Includes event log for full message history. The communication backbone. Triggers on: "inbox", "message", "send to", "broadcast", "check messages"

Authorize agent payments via Machine Payments Protocol (MPP). Handles microtransactions, API purchases, service subscriptions within budget governance. Triggered by pay, purchase, buy, transaction, payment, MPP, commerce.

Launch a specialized agent for a subtask. Assigns identity, workspace scope, available commands, and communication protocol. Supports tmux-based (visual) or subprocess (headless) execution backends. The primary way to parallelize work. Triggers on: "spawn", "launch agent", "start worker", "parallelize"

Autonomous iterative improvement loop. Agent modifies code, verifies against metrics, keeps improvements or reverts failures, and repeats. Uses git as memory — each change is committed, measured, and kept or discarded. Runs until a target metric is hit or max iterations reached. Triggers on: "autoresearch", "auto improve", "iterative improvement", "autonomous loop", "hill climb"

> Detect project type, compile, and report errors with actionable fixes.

Transform any GUI application into an agent-controllable CLI. 7-phase pipeline: analyze the GUI, design CLI commands, implement adapters, plan tests, write tests, document, and publish. Produces a standalone CLI tool that wraps GUI functionality for automation. Triggers on: "cli anything", "gui to cli", "make cli", "wrap gui", "automate application"

> Stage changes, generate a meaningful commit message, and commit.

> Create a pull request with summary, test plan, and proper labeling.

Guided spec creation through 9 phases from discovery to testing strategy. Supports spec types: library, feature, change. Produces a complete specification document with requirements, architecture, constraints, and YAML test data. Interactive or autonomous modes. Triggers on: "create spec", "write spec", "spec out", "specification", "design document"

> Systematic bug investigation: reproduce, isolate, hypothesize, fix, verify, prevent.

> Build, validate, and deploy to target environment.

> Run linters and formatters, report issues, auto-fix where possible.

> Restructure code for clarity and maintainability without changing behavior.

> Thorough code review across correctness, security, performance, and maintainability.

Enforces Test-Driven Development discipline with RED-GREEN-REFACTOR cycle

> Run test suite, report coverage, flag failures with diagnostics.

Human-in-the-loop approval gates. Creates approval requests for high-stakes actions like hiring agents, strategy proposals, or budget overrides. Tracks states from pending through approved/rejected/revision_requested. Blocks execution until resolved. Triggers on: "approve", "approval", "review request", "sign off"

3-tier budget enforcement for AI agent workspaces. Visibility dashboards (always on), soft alerts at 80% threshold, hard ceilings at 100% that auto-pause agents. Tracks per-agent, per-task, and per-project costs in both tokens and dollars. Triggers on: "budget", "cost", "spending", "token usage", "billing"