Tyler-R-Kendrick

Use when producing agent/LLM evals, synthetic simulation data, or self-improvement pipelines for prompts, code, skills, agents, harnesses, and workflows. Covers AgentEvals/AgentV, Agent Skills evals, ASSERT, GEPA, Trace, VISTA, Agent Lightning, SkillOpt, Simula-style data design, progressive disclosure, deterministic workspaces, and release evidence. USE FOR: eval creation, EVAL.yaml, AgentEvals, AgentV, evals.json, ASSERT, judge-traces, behavior taxonomy, judges, graders, rubrics, synthetic data, simulation data, Simula, QDC, source-grounded generation, prompt optimization, agent improvement, skill improvement, harness hardening, progressive disclosure, deterministic workflows, GEPA, Trace, VISTA, Agent Lightning, SkillOpt DO NOT USE FOR: ordinary unit/integration tests without AI quality criteria (use testing), refactoring without eval or trace feedback (use refactor), generic Agent Skills packaging without eval or improvement work (use agent-skills)

Use when working with AI agent protocols, standards, interoperability specifications, evaluation contracts, synthetic simulation data, improvement pipelines, and agent steering workflows. Covers MCP, A2A, ACP, Agent Skills, AGENTS.md, ADL, Improve, x402, AP2, MCP Apps, cagent, and learn. USE FOR: agent protocol selection, comparing MCP vs A2A vs ACP, understanding agent standards ecosystem, choosing payment protocols, choosing eval standards, choosing improvement techniques, choosing synthetic data simulation techniques, steering from user feedback DO NOT USE FOR: specific protocol, eval, or improvement implementation details (use the sub-skills: mcp, a2a, acp, improve, learn, x402, etc.)

Use when a user corrects, rejects, edits, or redirects an LLM/agent response and the correction should become a reusable reasoning strategy. Converts feedback into generalized learnings for ~/.agents/STEERING.md with linked RDF/Turtle evidence. USE FOR: user corrections, preference feedback, rejected agent behavior, reasoning strategy updates, steering file maintenance DO NOT USE FOR: storing task facts (use memory), ordinary skill authoring (use agent-skills), project instruction files unrelated to feedback (use agents-md)

Use when implementing the Agent-to-Agent (A2A) protocol for inter-agent communication, task delegation, and multi-agent collaboration. USE FOR: agent-to-agent communication, task delegation between agents, Agent Card publishing, multi-agent collaboration DO NOT USE FOR: tool integration (use mcp), agent payments (use ap2 or x402), agent definition (use adl)

Use when implementing the Agent Communication Protocol (ACP) for REST-based agent-to-agent communication, task delegation, and multimodal message exchange. USE FOR: ACP agent servers, ACP client integration, agent discovery via manifests, run lifecycle management, session-based stateful workflows, BeeAI agents DO NOT USE FOR: JSON-RPC agent communication (use a2a), tool integration for LLMs (use mcp), agent payments (use ap2 or x402), agent definition (use adl)

Use when defining AI agents declaratively with Agent Definition Language (ADL). Covers agent identity, LLM configuration, tools, permissions, RAG inputs, and governance metadata. USE FOR: declarative agent blueprints, agent identity and permissions, LLM configuration, governance metadata DO NOT USE FOR: agent runtime orchestration (use cagent), tool integration (use mcp), agent communication (use a2a)

Use when creating, packaging, or distributing Agent Skills. Covers the SKILL.md specification, frontmatter schema, naming conventions, marketplace publishing, and the skills-ref validator. USE FOR: creating SKILL.md files, packaging reusable agent capabilities, marketplace publishing, frontmatter schema validation DO NOT USE FOR: project-level agent guidance (use agents-md), agent runtime configuration (use adl or cagent)

Use when creating or updating AGENTS.md files to guide AI coding agents. Covers file structure, placement, content guidelines, and best practices for project-level agent instructions. USE FOR: project-specific agent instructions, build/test commands for agents, coding conventions, repository-level guidance DO NOT USE FOR: reusable cross-project skills (use agent-skills), agent runtime definition (use adl)

Improve a markdown prompt file using Agent Lightning APO (Automatic Prompt Optimization). Use when the user asks to optimize or improve a markdown prompt, or starts a message with /trainer-optimize.

Own the end-to-end trainer loop for agent contract targets (*.agent.md files, custom agent definitions, and agent instruction documents). Use this whenever the caller needs to research, synthesize datasets, optimize, validate, and write back a trained candidate for an agent-type target. Prefer this specialized loop whenever the selected target defines tool routing, MCP skill configuration, agent personas, or handoff behavior rather than raw prompts, code, or skill definitions.

Own the end-to-end trainer loop for Python code targets optimized with Microsoft Trace (nodes, bundles, models, and trainable agent components). Use this whenever the caller needs to research, synthesize test-based datasets, optimize, validate, and write back a trained candidate for a code-type target. Prefer this specialized loop for any Python file or callable that benefits from deterministic, test-based or benchmark-based feedback rather than open-ended language instruction quality.

Own the end-to-end trainer loop for Python code targets optimized with Microsoft Trace (nodes, bundles, models, and trainable agent components). Use this whenever the caller needs to research, synthesize test-based datasets, optimize, validate, and write back a trained candidate for a code-type target. Prefer this specialized loop for any Python file or callable that benefits from deterministic, test-based or benchmark-based feedback rather than open-ended language instruction quality.

Own the end-to-end trainer loop for prompt-like files (*.prompt.md, *.prompty, *.instructions.md, system prompts, and other natural-language instruction artifacts). Use this whenever the caller needs to research, synthesize datasets, optimize, validate, and write back a trained candidate for a prompt-type target. Prefer this specialized loop for any file whose primary content is natural-language instructions rather than code, skill configuration, or agent contracts.

Own the end-to-end trainer loop for prompt-like files (*.prompt.md, *.prompty, *.instructions.md, system prompts, and other natural-language instruction artifacts). Use this whenever the caller needs to research, synthesize datasets, optimize, validate, and write back a trained candidate for a prompt-type target. Prefer this specialized loop for any file whose primary content is natural-language instructions rather than code, skill configuration, or agent contracts.

Own the end-to-end trainer loop contract for a prompt-like file, skill contract, or agent contract after the caller has already chosen the concrete stage capabilities. Use this whenever the current agent must set up the local trainer workspace, coordinate stage sequencing, maintain workflow state, manage steering and candidates, recover from manual follow-up mode, and decide whether a trained candidate is safe to write back.

Own the end-to-end trainer loop for agent skill targets (SKILL.md files and their supporting references, scripts, and evals). Use this whenever the caller needs to research, synthesize datasets, optimize, validate, and write back a trained candidate for a skill-type target. Prefer this specialized loop whenever the selected target is a SKILL.md file or the user wants to improve skill triggering accuracy, body content quality, or progressive-disclosure structure.

Use when coordinating, inspecting, running, or verifying the Project Khepri modernization workflow, Microsoft Agent Framework workflow, GHCP SDK custom-agent registry, increment squad workflow, AgentEvals gates, or app/data/infra/security modernization sequence.

Deterministic workflow for GitHub Copilot modernization agents.

{what this skill teaches agents}

TDD loop for generating SDK-first squads with AgentEvals, evaluators, test data, rubrics, and live-evals.

Use when architecture-affecting code, agent profiles, workflow contracts, hooks, skills, MCP configuration, evals, CI, repository structure, or documentation changes could make Project Khepri docs or Mermaid diagrams stale.

Use this skill when the user corrects an agent, says prior behavior was wrong, asks the agent to remember a preference, or gives steering that should prevent repeat mistakes. Capture a succinct generalized correction in STEERING.md for all Project Khepri agents.

Discovery checklist for scoping modernization before implementation.

Planner and evolution handshake for generating a modernization-specific squad.

Runs scoped research as a composite skill through the shared skill registry and router.

Create and maintain paper research packets in this repo with the required research/<paper>/ layout, architecture docs, and experiment implementations aligned to the agent-browser TypeScript stack. Use this whenever the user asks to add a paper, summarize research, create experiment plans, or implement a paper capability as a reference architecture.

ARIS-inspired deep research workflow with adversarial executor/reviewer loops, persistent memory, and claim-evidence assurance checks.

Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.

Operating guide for the agent-browser workspace shell. Use this whenever the user asks how to inspect, navigate, or modify the active agent-browser workspace, Files surface, browser pages, sessions, clipboard history, render panes, or WebMCP tool flows. Prefer it before improvising tool chains because the active workspace, workspace files, and mounted session drives have specific semantics in this project.

Create an AgentEvals-style eval suite for a named agent under .agents/<agent-name>/.evals/. Use this whenever the user asks for an eval, regression suite, benchmark, or repeatable acceptance test for a workspace agent. Prefer it even when the user asks for a smoke test or acceptance check without naming AgentEvals directly.

Create a scoped agent folder with an AGENTS.md file under .agents/<agent-name>/. Use this whenever the user asks for a new agent, reusable agent instructions, a workspace-scoped AGENTS.md, or a named automation persona inside the current workspace. Prefer it even when the user only describes the role and not the file layout.

Create a reusable agent skill bundle under .agents/skills/<skill-name>/ that follows agentskills.io conventions. Use this whenever the user asks for a SKILL.md, reusable workflow skill, skill scaffold, or packaged agent capability inside the current workspace. Prefer it even when the user only describes the capability and not the folder structure.

Configure and use reputable web scraping for TK TechNews. Use when an agent needs Firecrawl MCP, Firecrawl-backed scraping, dynamic page extraction, web source troubleshooting, or guidance on when to use RSS, local scraping, Firecrawl, or YouTube transcripts for cited article generation.

Use the local TK TechNews YouTube Data API MCP server. Trigger when an agent needs YouTube channel info, playlist info, playlist videos, video metadata, YouTube search, caption track listing, caption downloads, or VS Code MCP setup for the repo's local YouTube server.

Use the local TK TechNews MCP server wrapping jdepoix/youtube-transcript-api. Trigger when an agent needs YouTube transcripts, generated subtitles, transcript language discovery, translation, SRT/VTT/text/JSON transcript output, or VS Code MCP setup for transcript extraction without a YouTube Data API key.

Draft cited TK TechNews articles from the normalized source summary ledger. Use when an agent needs to turn data/summaries/latest.json into a Markdown article, preserve citations in frontmatter, or create an initial explainer draft for editorial refinement.

Run the durable TK TechNews source-to-article pipeline. Use when an agent needs to ingest one URI, create a cited source brief, enrich it into the temporal knowledge graph, aggregate enriched docs by day, or generate an Astro article from an aggregate brief.

Validate and publish TK TechNews static Astro content. Use when an agent needs to check citation integrity, run the Astro build, preview the site, or prepare a generated article for commit or deployment.

Collect technology-news source material for TK TechNews. Use when an agent needs to fetch RSS feed items, scrape configured web pages, pull YouTube transcripts, refresh data/summaries/latest.json, or prepare source summaries before drafting an article in this repository.

Create, load, merge, and distribute persistent artifact plugin bundles for activation steering. Use this skill when the user wants to manage steering artifacts, create distributable plugin packs, load model bundles, merge multiple plugins, write new artifact directories, or work with the plugin directory tree. Also trigger when the user mentions artifact plugins, plugin bundles, plugin manifests, controllers.json, activations.json, or the artifacts/ directory layout.

Define, discover, and manage cognitive feature specifications and steering vectors for LLM activation steering. Use this skill when the user wants to create feature specs, define extraction examples with positive/negative labels, set evaluation criteria, discover feature vectors from specs, persist discovered vectors, work with the standard feature catalog, or manage the feature lifecycle. Also trigger when the user mentions feature extraction, feature catalog, feature specs, cognitive features, or interaction features.

Use when creating, compiling, validating, running, or debugging GitHub Agentic Workflows with the gh-aw CLI.

Build and query Neo4j-backed reasoning trajectory graphs for the hybrid meta-cognition agent. Use this skill when the user wants to persist task plans, subgoals, constraints, run states, and verifier outcomes in a Neo4j graph, retrieve evidence paths with PathRAG, or store drift-correction records. Also trigger when the user mentions Neo4j, GraphRAG, PathRAG, reasoning trajectories, graph store, evidence paths, or task plan graphs — even if they don't say "graphrag" explicitly.

Build and run a hybrid planner/retriever/steered-executor/verifier meta-cognition agent loop with persistent steering memory. Use this skill when the user wants to create an agent that plans tasks, retrieves context, applies activation steering, verifies results, and writes back to memory. Also use when the user mentions HybridMetaCognitionAgent, SteeredExecutor, planner/verifier loops, steering controllers, agent runs, or activation traces — even if they don't say "hybrid agent" explicitly.

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

Run activation-steering feature discovery for a Hugging Face model. Use this skill whenever the user says /steering, wants to steer a model, extract a cognitive feature, generate steering vectors, run feature discovery, or produce steering artifacts. Specify a model (default: gpt2) and an optional feature name. If a feature is supplied, generate inputs, expected outputs, then run an extraction pass and output artifacts. If no feature is supplied, auto-pick one that hasn't already been extracted. If no test data is supplied, generate synthetic examples; otherwise use the user's data and fill in whatever's missing.

Use the speculate skill when you need critique-ready hypotheses from new observations: recall context, run the mutation pipeline on a fresh observation, retrieve ranked inference and facet results from an inference/* branch, and present candidate text, provenance, assumptions, uncertainty, and review priority without promoting anything to trusted memory.

Use the discover skill to find connections between claims through facet relations and to rank inference and facet candidates using manifold geometry. Covers facet modeling, manifold scoring, relatedness, distance, uncertainty, and retrieval of discovery results.

Use the infer skill to generate speculative inference candidates from claims, persist them on branch-local inference graphs, and run the full mutation pipeline that coordinates active memory, claims, Terminus persistence, and speculative ranking.

Use the memorize skill to store observations, entities, tasks, and claims into the filesystem-backed active memory. Covers session lifecycle, working-set writes, entity/task card creation, claim extraction, and candidate promotion.

Use the recall skill to retrieve composed context from active memory, trusted Terminus graphs, and optional speculative branches. Also covers journal-backed history of past working-memory mutations.

Use the reflect skill to persist memories, claims, and inference nodes into the TerminusDB temporal graph. Covers branch lifecycle, trusted session writes, temporal recall, and the separation between session and inference branches.

Use when initializing Universal Token Killer schema artifacts for a project, discovering registered Copilot tools, seeding .utk output schemas, or narrowing schema generation to specific tool ids with descriptions or sample outputs

Use when installing, operating, recovering, or extending Universal Token Killer for GitHub Copilot tool-hook mediation, .utk artifacts, serializer configuration, schema routing, compact tool-output recovery, or RTK-parity token optimization

Use when compressing prompts, agent skills, AGENTS.md files, GitHub Copilot custom agents, subagent definitions, or bulky artifacts before sending text to LLM context; includes detoks-prompt CLI/MCP use, skill consolidation, AGENTS.md memory/skill extraction, and GHCP subagent/frontmatter orchestration

Use when configuring, testing, or documenting UTK model-proxy behavior for GitHub Copilot context optimization.

Use when optimizing Agent Skills for lower token usage, creating compact companion skills, splitting skill instructions into references, extracting deterministic workflow scripts, or adding eval-backed skill compression

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

Use this skill whenever the user wants to create, author, or generate timed speech scripts, audio dubbing scripts, narration tracks, or any kind of synchronized audio content using the ReDub library. Triggers include: "generate a dubbing script", "create timed narration", "write a TTML/DAPT script", "author a speech document with timing", "ReDub", "timed audio", "dubbing with agents", or any request to produce synchronized spoken audio with timing metadata. Also use when the user wants to understand how to structure audio timing, assign lines to voice actors/agents, add pronunciation hints, or convert frame-based timecodes to TTML seconds.

Use, extend, test, and integrate Epoch's TypeScript DVCS, CLI, WASM, and platform packages.

Install and use GitHub Agentic Workflows (gh-aw) to create, compile, run, debug, and manage AI-powered workflows written in Markdown. Use when asked to create a new agentic workflow, compile a workflow to GitHub Actions YAML, trigger or monitor a workflow run, or work with any file under .github/workflows/*.md.

Use this skill for high-assurance validation of reasoning, claims, and proof sketches with the repository's full workflow or direct SMT/Lean checks.