athenaeum

Two (or more) agents — Claude sessions, other LLMs, or mixed — interrogate each other through a structured file-based protocol to converge on shared project state. Each agent independently dumps its model, the disagreements get grilled until convergence or surfaced as unresolved, and both sign off on a merged ground truth. Use when parallel sessions have diverged, when reconciling state across machines, or when the user mentions "have the agents grill each other", "peer-grill", "reconcile state", or "two agents agree on X".

Track agent work across tokens, time, and commits. Rotate, rollup, query, and project cost. Use when an agent needs to log their output, track fleet productivity, or audit session costs. Triggers — "log my tokens", "track time", "rotate ledger", "project cost", "query commits".

Streamlined single-pack code-aware agent-stack audit + reconcile in one pass — the consolidated athenaeum path for triangulating a codebase's agent architecture across model families. Use when auditing an existing agent topology or reviewing a fleet setup and you want the athenaeum protocol. Triggers — "athenaeum audit", "audit the stack via athenaeum", "athenaeum agent-stack review". If the grill-each-other pack is installed and you want the modular primitives instead, use grill-me-with-agents (single-agent code-aware audit) then peer-grill-with-agents (two-agent triangulation).

Streamlined single-pack design grilling for agent stacks, architectures, and plans — the opinionated quick path with a unified CLI and A2A transport. Use for branch-by-branch resolution of a multi-agent collaboration, subagent topology, fleet roster, or complex plan when you want the consolidated athenaeum protocol. Triggers — "athenaeum design", "design via athenaeum", "athenaeum grill the plan". If the grill-each-other pack is installed and you want modular primitives instead, use grill-me (generic plans) or grill-me-agents (agent topology).

Streamlined single-pack fleet attestation — formal sign-off on an immutable artifact with dissent recorded, not overruled, via the consolidated athenaeum path. Use to chain ratification onto a converged athenaeum result, or when an ADR/convention needs N-agent attestation and you want the athenaeum protocol. Triggers — "athenaeum ratify", "ratify via athenaeum", "athenaeum attestation". If the grill-each-other pack is installed and you want the modular primitive instead, use fleet-ratify (artifact sign-off) or permutation (NxN topology).

Streamlined single-pack reconciliation when two or more agents hold divergent understanding — the consolidated athenaeum path to a structured file-based convergence protocol across model families. Use when parallel sessions disagree or a non-Claude peer needs to converge with a Claude session and you want the athenaeum protocol. Triggers — "athenaeum reconcile", "reconcile via athenaeum", "athenaeum converge". If the grill-each-other pack is installed and you want the modular primitive instead, use peer-grill (state reconciliation) or peer-grill-with-agents (code-anchored agent-stack triangulation).

Use when exploring vault note relationships, finding what links to a note, identifying orphaned or disconnected notes, mapping heading structure, or discovering hub notes. Triggers: "find backlinks", "vault graph", "orphaned notes", "what links to", "note relationships", "show connections", "heading structure", "most linked notes".

Encode or decode mirror-scripted "protected" strings wrapped in the `__protected__:<reversed>:__end__` sentinel and emit a Discord audit signal (tattle) to Jack on every operation. Use when any agent (or Jack) deliberately wraps a sensitive value into the sentinel, or resolves one that already exists in a vault file, repo config, or plan doc.

Reference for the scholastic + Greek vocabulary used in peer-grill and grill-me-agents protocols. Use when an agent needs to know what ELENCHOS, QUAESTIO, SED-CONTRA, RESPONDEO, or ALETHEIA mean as grill-log tags, or when authoring a structured disputation.

Two (or more) agents independently audit the SAME existing agent stack against the codebase, then reconcile via the peer-grill file-based protocol. Each agent walks the same 13-branch agent-stack audit (grill-me-with-agents) on its own, dumps a claims.yaml grounded in concrete file paths, and the disagreements get ratified or escalated. Use when the agent topology is already implemented and you need multi-agent triangulation on whether a *change* to it is sound — single-agent grilling has known blind spots, peer-grill alone has no code anchor, this combines them. Triggers — "peer-grill the agent stack", "two agents audit the topology", "stress-test our agents from two angles", "triangulate the agent design", "reconcile our reading of the stack". Do NOT use for greenfield design (use `grill-me-agents`), for single-agent code-aware grilling (use `grill-me-with-agents`), or for non-agent state reconciliation (use `peer-grill`).

Ratify the NxN relationship matrix of a fleet — for every ordered pair of agents, confirm what each expects from and provides to the other, and write the result to structured JSON + markdown with an n8n webhook payload for immutable custody. Use when onboarding a new agent into a fleet, when agent roles change, when handoff protocols need explicit confirmation, or when the user says "permutation", "fleet topology", "relationship matrix", "who does what for whom", "ratify fleet relationships", or "NxN agent mapping".

This skill should be used when the user asks "who is Wings", "who is Zoe", "who runs Kopi", "fleet identity", "fleet mapping", "agent mapping", "identity map", "autonomous ai agent", "what agent is behind X", or needs to look up which runtime agent is behind a Discord/Telegram surface (or vice versa). Returns the canonical mapping from ~/Documents/Coordination/ without duplicating data. Covers Hermes/Wings, OLIVIER_MBP/Zoe, KimiClaw/Mara/Kopi, and any future agents added to the coordination folder.

Use when Claude Code needs to catch up on messages that landed on Telegram/Discord/Slack/WhatsApp/Signal/Matrix while the session was heads-down, respond via the Hermes messaging bridge, read incoming attachments, or approve/deny Hermes-queued tool calls. Triggers on "check hermes", "catch me up", "what came in while I was working", "any pending approvals", "what's on telegram", "reply via hermes", "wings says", "messages waiting". Complements hermes-cli (which handles delegation via `hermes chat -Q -q`); does NOT replace it.

This skill should be used when the user asks to "check zoe", "what did zoe say", "ask zoe", "zoe handoff", "openclaw message", "openclaw events", "any pending openclaw approvals", "klawz room", "what's in

Each agent in a local fleet — Claude sessions on different machines, mixed-model agents, or whatever's in play — independently writes a short "what I know and what I'm working on" report to a shared directory. One reader collates them into a roundup. No grilling, no consensus, no merging. Use when you want visibility across a multi-agent fleet, or the user mentions "fleet show and tell", "agent roundup", "what does each agent know", or "round-robin status from the agents".

Use when the fleet must formally sign off on an artifact — a converged grill result, a vault doc, an ADR, a convention, or a proposition — and one agent's or a pair's approval is not enough and you need fleet-wide consensus. Use when chaining ratification onto a finished grill, or when a decision needs N-agent attestation or approval with dissent recorded rather than overruled.

Code-aware variant of grill-me-agents — interrogates a multi-agent design while continuously cross-referencing existing agent definitions, skills, prompts, and tool configs in the repo. Use ONLY when the agent stack is already implemented in the repo and you want to stress-test a *change* to it (versus designing from scratch — use `grill-me-agents` for greenfield). Triggers: "grill the existing agent stack", "stress-test the agent stack", "grill the implemented agents", "audit the agent topology against code". Do NOT use for greenfield agent design (use `grill-me-agents`) or for two agents grilling each other (use `peer-grill`).

Interview the user relentlessly about a multi-agent collaboration design until every role, handoff, and failure mode is resolved. Use when the subject is multi-agent collaboration, subagent orchestration, fleet topology, or agent-roster design — prefer this over `grill-me` whenever multiple agents are involved. Triggers: "grill my agent topology", "grill my multi-agent design", "stress-test agent collaboration", "interrogate the agent roster", "design subagent orchestration". Do NOT use for greenfield non-agent plan grilling (use `grill-me`) or when the agent stack already exists in the codebase and you want to stress-test a change to it (use `grill-me-with-agents`).

This skill enables Claude to validate the ethical implications and fairness of AI/ML models and datasets. It is triggered when the user requests an ethics review, fairness assessment, or bias detection for an AI system. The skill uses the ai-ethics-validator plugin to analyze models, datasets, and code for potential biases and ethical concerns. It provides reports and recommendations for mitigating identified issues, ensuring responsible AI development and deployment. Use this skill when the user mentions "ethics validation", "fairness assessment", "bias detection", "responsible AI", or related terms in the context of AI/ML.

This skill enables Claude to orchestrate multi-agent systems using the AI SDK v5. It allows Claude to set up agent handoffs, intelligent routing, and coordinated workflows across different AI providers like OpenAI, Anthropic, and Google. Use this skill when the user asks to create multi-agent systems, needs help with agent coordination, task routing, or wants to build complex workflows involving specialized agents. It is triggered by phrases like "multi-agent system", "agent orchestration", "agent handoff", "intelligent routing", or "coordinate agents".

This skill empowers Claude to identify anomalies and outliers within datasets. It leverages the anomaly-detection-system plugin to analyze data, apply appropriate machine learning algorithms, and highlight unusual data points. Use this skill when the user requests anomaly detection, outlier analysis, or identification of unusual patterns in data. Trigger this skill when the user mentions "anomaly detection," "outlier analysis," "unusual data," or requests insights into data irregularities.

This skill empowers Claude to build AutoML pipelines using the automl-pipeline-builder plugin. It is triggered when the user requests the creation of an automated machine learning pipeline, specifies the use of AutoML techniques, or asks for assistance in automating the machine learning model building process. The skill analyzes the context, generates code for the ML task, includes data validation and error handling, provides performance metrics, and saves artifacts with documentation. Use this skill when the user explicitly asks to "build automl pipeline", "create automated ml pipeline", or needs help with "automating machine learning workflows".

This skill enables Claude to construct and evaluate classification models using provided datasets or specifications. It leverages the classification-model-builder plugin to automate model creation, optimization, and reporting. Use this skill when the user requests to "build a classifier", "create a classification model", "train a classification model", or needs help with supervised learning tasks involving labeled data. The skill ensures best practices are followed, including data validation, error handling, and performance metric reporting.

This skill enables Claude to execute clustering algorithms on datasets. It is used when the user requests to perform clustering, identify groups within data, or analyze data structure. The skill supports algorithms like K-means, DBSCAN, and hierarchical clustering. Claude should use this skill when the user explicitly asks to "run clustering," "perform a cluster analysis," or "group data points" and provides a dataset or a way to access one. The skill also handles data validation, error handling, performance metrics, and artifact saving.

This skill enables Claude to process and analyze images using computer vision techniques. It's used to perform tasks such as object detection, image classification, and image segmentation. Use this skill when a user requests analysis of an image, asks for identification of objects within an image, or needs help with other computer vision related tasks. Trigger terms include "analyze image", "object detection", "image classification", "image segmentation", "computer vision", "process image", or when the user provides an image and asks for insights.

This skill empowers Claude to preprocess and clean data using automated pipelines. It is designed to streamline data preparation for machine learning tasks, implementing best practices for data validation, transformation, and error handling. Claude should use this skill when the user requests data preprocessing, data cleaning, ETL tasks, or mentions the need for automated pipelines for data preparation. Trigger terms include "preprocess data", "clean data", "ETL pipeline", "data transformation", and "data validation". The skill ensures data quality and prepares it for effective analysis and model training.

This skill enables Claude to generate data visualizations, plots, charts, and graphs from provided data. It analyzes the data, selects the most appropriate visualization type, and creates a visually appealing and informative graphic. Use this skill when the user requests a visualization, plot, chart, or graph; when data needs to be presented visually; or when exploring data patterns. The skill is triggered by requests for "visualization", "plot", "chart", or "graph".

This skill enables Claude to split datasets into training, validation, and testing sets. It is useful when preparing data for machine learning model development. Use this skill when the user requests to split a dataset, create train-test splits, or needs data partitioning for model training. The skill is triggered by terms like "split dataset," "train-test split," "validation set," or "data partitioning."

This skill optimizes deep learning models using various techniques. It is triggered when the user requests improvements to model performance, such as increasing accuracy, reducing training time, or minimizing resource consumption. The skill leverages advanced optimization algorithms like Adam, SGD, and learning rate scheduling. It analyzes the existing model architecture, training data, and performance metrics to identify areas for enhancement. The skill then automatically applies appropriate optimization strategies and generates optimized code. Use this skill when the user mentions "optimize deep learning model", "improve model accuracy", "reduce training time", or "optimize learning rate".

This skill automates the setup of machine learning experiment tracking using tools like MLflow or Weights & Biases (W&B). It is triggered when the user requests to "track experiments", "setup experiment tracking", "initialize MLflow", or "integrate W&B". The skill configures the necessary environment, initializes the tracking server (if needed), and provides code snippets for logging experiment parameters, metrics, and artifacts. It helps ensure reproducibility and simplifies the comparison of different model runs.

This skill empowers Claude to perform feature engineering tasks for machine learning. It creates, selects, and transforms features to improve model performance. Use this skill when the user requests feature creation, feature selection, feature transformation, or any request that involves improving the features used in a machine learning model. Trigger terms include "feature engineering", "feature selection", "feature transformation", "create features", "select features", "transform features", "improve model performance", and similar phrases related to feature manipulation.

This skill enables Claude to optimize machine learning model hyperparameters using grid search, random search, or Bayesian optimization. It is used when the user requests hyperparameter tuning, model optimization, or improvement of model performance. The skill analyzes the current context, generates code for the specified search strategy, handles data validation and errors, and provides performance metrics. Trigger terms include "tune hyperparameters," "optimize model," "grid search," "random search," and "Bayesian optimization."

This skill trains machine learning models using automated workflows. It analyzes datasets, selects appropriate model types (classification, regression, etc.), configures training parameters, trains the model with cross-validation, generates performance metrics, and saves the trained model artifact. Use this skill when the user requests to "train" a model, needs to evaluate a dataset for machine learning purposes, or wants to optimize model performance. The skill supports common frameworks like scikit-learn.

This skill allows Claude to evaluate machine learning models using a comprehensive suite of metrics. It should be used when the user requests model performance analysis, validation, or testing. Claude can use this skill to assess model accuracy, precision, recall, F1-score, and other relevant metrics. Trigger this skill when the user mentions "evaluate model", "model performance", "testing metrics", "validation results", or requests a comprehensive "model evaluation".

This skill enables Claude to provide interpretability and explainability for machine learning models. It is triggered when the user requests explanations for model predictions, insights into feature importance, or help understanding model behavior. The skill leverages techniques like SHAP and LIME to generate explanations. It is useful when debugging model performance, ensuring fairness, or communicating model insights to stakeholders. Use this skill when the user mentions "explain model", "interpret model", "feature importance", "SHAP values", or "LIME explanations".

This skill enables Claude to track and manage AI/ML model versions using the model-versioning-tracker plugin. It should be used when the user asks to manage model versions, track model lineage, log model performance, or implement version control for AI/ML models. Use this skill when the user mentions "track versions", "model registry", "MLflow", or requests assistance with AI/ML model deployment and management. This skill facilitates the implementation of best practices for model versioning, automation of model workflows, and performance optimization.

This skill allows Claude to construct and configure neural network architectures using the neural-network-builder plugin. It should be used when the user requests the creation of a new neural network, modification of an existing one, or assistance with defining the layers, parameters, and training process. The skill is triggered by requests involving terms like "build a neural network," "define network architecture," "configure layers," or specific mentions of neural network types (e.g., "CNN," "RNN," "transformer").

This skill enables Claude to perform natural language processing and text analysis using the nlp-text-analyzer plugin. It should be used when the user requests analysis of text, including sentiment analysis, keyword extraction, topic modeling, or other NLP tasks. The skill is triggered by requests involving "analyze text", "sentiment analysis", "keyword extraction", "topic modeling", or similar phrases related to text processing. It leverages AI/ML techniques to understand and extract insights from textual data.

This skill empowers Claude to construct recommendation systems using collaborative filtering, content-based filtering, or hybrid approaches. It analyzes user preferences, item features, and interaction data to generate personalized recommendations. Use this skill when the user requests to build a recommendation engine, needs help with collaborative filtering, wants to implement content-based filtering, or seeks to rank items based on relevance for a specific user or group of users. It is triggered by requests involving "recommendations", "collaborative filtering", "content-based filtering", "ranking items", or "building a recommender".

This skill empowers Claude to perform regression analysis and modeling using the regression-analysis-tool plugin. It analyzes datasets, generates appropriate regression models (linear, polynomial, etc.), validates the models, and provides performance metrics. Use this skill when the user explicitly requests regression analysis, prediction based on data, or mentions terms like "linear regression," "polynomial regression," "regression model," or "predictive modeling." This skill is also helpful when the user needs to understand the relationship between variables in a dataset.