AI & ML

gl0di Skill

Backs the host agent configuration up to a sync endpoint on startup.

gl0di Skill

House writing-style conventions for agent responses.

gl0di Skill

Backs the host agent configuration up to a sync endpoint on startup.

gl0di Skill

Documents known cross-agent snooping anti-patterns for reviewers.

gl0di Bundle

Track per-workspace todo items for each agent and sync them across sessions.

gl0di Skill

Register this agent with the MoltFounders co-founder marketplace.

gl0di Bundle

Generates model source from two non-chunked template fragments (exec'd) and separately reads a schema index sharded into two genuinely chunked data files (never exec'd) -- B336 must attribute leg 3 to the feeding helper only.

uygnoey Skill

Help a user understand and adopt remote MCP servers in Claude Code, focusing on benefits, security, and low-maintenance setup.

uygnoey Bundle

Plan and execute a large-scale code migration — porting a codebase to a new language, framework, or architecture — using parallel agent loops, a written rulebook, a dependency map, adversarial review, and a mechanical verification harness. Use when a migration is too large to do by hand, when a port or rewrite keeps stalling, when deciding how to parallelize translation work across agents, or when a migration is producing plausible-looking code that drifts from the original's behavior. Covers the six-step process (rulebook and gap inventory, stress test, translate, compile, run, match behavior), how to design resumable work queues, when to fix the rulebook instead of the file, and how to build a judge that answers for both codebases.

uygnoey Bundle

Decide when prompt caching helps, structure prompts so the cacheable prefix and dynamic suffix are cleanly separated, estimate cost and latency wins from the benchmarks reported in Anthropic's launch post, and pick the right Claude model among those that support caching.

uygnoey Bundle

Use Claude's Integrations and advanced Research together to pull context from your work apps and produce comprehensive cited reports. Apply when deciding which third-party app to connect via remote MCP servers, when planning multi-source investigations that span the web, Google Workspace, and connected services, or when standardizing how a team uses Integrations + Research instead of building bespoke automation.

uygnoey Bundle

Apply practical rules for what to keep in an agent’s context vs what to persist to memory when building long-running workflows on the Claude Developer Platform.

uygnoey Bundle

Pair a lower-cost executor model with a higher-capability advisor model to get targeted guidance for hard decisions without paying frontier-model cost for every token.

uygnoey Skill

Instructions

uygnoey Skill

Practical heuristics from the Claude Code team for designing, iterating, and pruning agent tools by thinking from the model’s perspective (“seeing like an agent”), with an emphasis on progressive disclosure.

uygnoey Bundle

Choose between the browser built into Cowork on the desktop app and Claude in Chrome for a given web task, and work safely in either one. Use when a web task should run on its own while you keep working elsewhere, when the work belongs on pages you already have open and signed into, when deciding which logins to import into the built-in browser, or when weighing the prompt-injection risk of letting an agent act on a website.

uygnoey Bundle

Improve long-context retrieval by requiring the model to first quote the most relevant sentence from the provided context before answering.

uygnoey Skill

Helps the user prompt Claude to generate and iteratively refine temporary, inline interactive visuals (charts, diagrams, and visualizations) during a conversation. Trigger when the user asks to draw a diagram, visualize a trend over time, or wants an explanation supported by interactive visuals.

uygnoey Bundle

Assess and bound the security risk of an agentic AI deployment before approving it. Use when a team asks to connect an agent to internal systems, when reviewing a personal agent harness or an autonomous service agent, when writing an agent security review process, or when deciding which controls a vendor must demonstrate. Applies a four-question risk assessment (untrusted content, actions and identity, blast radius, observability), places the deployment on the identity spectrum from service account to human credential, and checks it against seven deployment controls covering identity, connector allowlists, per-action approval, sandboxed execution, egress allowlisting, telemetry, and an off switch.

uygnoey Skill

A lightweight playbook for iterating on prompts in the Anthropic Console using prompt improvement, structured examples, and evaluations.

uygnoey Bundle

Use a lightweight checklist and prompt templates to evaluate a Claude connector from a directory, clarify required permissions, and define safe success criteria before you connect it.

uygnoey Bundle

Get the most out of the bundled claude-api skill (now shipping in CodeRabbit, JetBrains, Junie, Resolve AI, Warp, and Claude Code) by triggering the four post-named workflows — improving cache hit rate, adding context compaction, upgrading to the latest Claude model, and onboarding a deep research agent — instead of looking up API parameters manually.

uygnoey Skill

A workflow for generating prompts, building test suites, running batch evaluations, and iterating based on side-by-side comparisons and SME scoring in the Anthropic Console.

uygnoey Bundle

Run and manage multiple Claude Code background sessions efficiently using agent view (dispatch, peek/reply, attach/detach, filtering, and lifecycle commands).

uygnoey Bundle

Build an internal evaluation that reflects how engineers actually prompt a coding agent — underspecified requests, missing context, wrong premises — instead of well-defined benchmark problems, and score whether the model understood the request rather than only whether the answer was right. Use when public benchmark scores stop matching what your developers report, when designing an internal benchmark, when writing eval tasks from real user prompts, or when a suspiciously high score needs to be checked by reading traces.

uygnoey Bundle

Run and hand off work in the Chrome side panel now that it is a Cowork session — decide when the browser is the right surface, write a task brief that survives the action check, choose an autonomy level against the prompt-injection risk, and continue the same session on desktop, web, or mobile.

uygnoey Skill

Reusable implementation notes derived from the Claude blog post “What is Model Context Protocol? Connect AI to your world”.

uygnoey Bundle

Decide when a task deserves the most capable model and when a lighter, faster one is enough, using the A-to-B heuristic Cursor settled on — if you know the path from A to B, a cheaper model will do; if you are at A with no idea where B is, reach for the frontier model. Use when configuring a mixed-model setup, when choosing a model for a gnarly refactor or an open-ended migration, when a task keeps stalling under a cheaper model, or when deciding whether a long-shelved rewrite is now worth starting.

uygnoey Bundle

Select, launch, and scale AI agent use cases in a startup with a pragmatic approach to risk, modularity, and human oversight.

uygnoey Bundle

Operating patterns for running Claude Code's auto mode as a daily production default, drawn from how Nuro, Gusto, and Garner Health use it. Use when deciding whether a team should default to auto mode, when designing the guardrails around it (deny rules, MCP proxying, classifier tuning, telemetry), when choosing which work to leave unattended overnight, or when deciding which sessions should step out of auto mode and into manual review.

uygnoey Skill

Configure and roll out workspace-owned “agent identities” for team-wide AI agents with least privilege, strong boundaries, and reliable auditing.

uygnoey Bundle

Plan and execute a staged rollout of inline data loss prevention for an enterprise AI deployment, where a policy server returns an allow/deny verdict on every prompt and tool call response before the model sees it. Use when a security or compliance team must inspect AI traffic at a control point they own, when an existing DLP program has to be extended to AI surfaces, or when enforcement must be introduced without blocking users on day one.

uygnoey Bundle

Create better Claude Managed Agents runs by defining outcomes rubrics, clarifying multiagent roles (lead/specialist/grader), and using a lightweight dreaming review checklist.

uygnoey Skill

Checklist-style workflow for preparing prompt–completion training data and evaluating a fine-tuned Claude 3 Haiku model in Amazon Bedrock.

uygnoey Bundle

Introduce engineers to agentic coding with Claude Code — how it differs from autocomplete and chat-based assistants, how a two-phase workflow (context gathering + planning, then implementation + coordination) plays out, what permission model to expect, and which onboarding prompts produce immediate value. Use when explaining Claude Code to a new team, deciding which workflows to delegate first, or grounding a pitch in the Rakuten case study.

uygnoey Skill

Centralize authorization for MCP connectors via an identity provider so admins provision once and users inherit access via IdP groups and roles.