Browse reusable Agent Skills, each with a clear purpose and practical guidance.
Use Inspect Scout to analyze agent trajectories from evaluation log files. Runs default and custom scanners to detect external failures, formatting issues, reward hacking, and ethical refusals. Use when user asks to check/analyze agent trajectories. Trigger when the user asks you to run the "Check Agent Trajectories" workflow.
Create an evaluation report for a README by selecting models, estimating costs, running evaluations, and formatting results tables. Use when user asks to make/create/generate an evaluation report. Trigger when the user asks you to run the "Make An Evaluation Report" workflow.
Benchmark and compare LLM models for SUB/WAVE's on-air calls — track picks, segments, listener requests, DJ scripts, banter, and programme beats — in both candidate-pool and agent modes, using controller/scripts/llm-bench (the matrix harness) or the legacy picker-only controller/scripts/picker-test.mjs. Use this skill whenever the user wants to assess, benchmark, compare, or test which LLM model to run the station on — phrases like "which model should I use", "benchmark the picker / the DJ / this model", "test these models", "is <model> good enough for the radio", "compare the ollama models", "will a small model work", "run llm-bench", "run picker-test", or when diagnosing slow/failing djAgentPick / djAgentSegment / generate* calls and the model choice is suspect. Trigger it even if the user doesn't name a harness — any request to evaluate model reliability or choose a model for SUB/WAVE belongs here. This skill only measures and recommends; it does NOT change the live station's configured model.
Customise the prd-taskmaster plugin workflow via curated brainstorm questions. The AI asks, the user answers in plain English, and the skill writes their preferences to .atlas-ai/config/atlas.json. Future runs of prd-taskmaster read that file and apply user preferences to phase gates, validation strictness, default provider, preferred execution mode, and template choice. For deeper tweaks beyond the curated questions, users can hand-edit files in .atlas-ai/customizations/. Use when the user says "customise workflow", "customize workflow", "adjust my PRD settings", "tune the skill", or wants to change how prd-taskmaster behaves.
What to do when a tool fails, an API hits a rate limit, or a site blocks scraping. Maps every primary tool to its best fallback so a single failure doesn't block the search.
Detailed 5-phase workflow for creating agents - from requirements analysis through validation and iterative refinement
5-layer testing approach for agent validation including adversarial testing, security validation, and prompt injection resistance
Detailed review process, v2 validation checklist, and scoring methodology for agent definition reviews
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
How the nWave buddy agent reads a project to answer questions — detection, order of inspection, and citation discipline.