SkillsKit

Company: SkillsKit

Self-reported basis: The description is from the author's own submission to the OpenAI 2026 hackathon on Devpost. It is unverified and self-reported.

What it appears to be: A tool that converts research into reusable, eval-first agent skills using AI-assisted workflows. It operates within a pipeline designed for agent-native development, with an emphasis on validation and reusability.

What changed: The author states they were frustrated by losing knowledge after deep research and wanted to make research reusable as agent skills. This led to the creation of SkillsKit during a hackathon.

Single most important open question: Is there evidence that this tool is being used beyond the author's own workflow, or has it achieved any adoption or traction outside of its creator?

The description states that SkillsKit is a system for turning research into reusable agent skills. It includes:

• A core skill /skill-from-research that takes a research pack (a folder of reports and notes) and inventories it.
• Verification of claims against primary sources.
• Splitting material into single-purpose skills.
• Eval-first authoring: each skill ships with should-trigger and should-not-trigger evals.
• Tools for scaffolding repositories (/create-skill-repo), publishing (/publish-repo), and authoring from ideas (/add-skill).
• Output is installable via npx skills add <repo> or directly within agent environments using the Agent Skills standard.

Inference: The system is built to be agent-native, with a focus on eval-first design and CI validation. It uses AI tools like GPT-5.6 and Codex in its pipeline.

Not evidenced: No information about actual usage, customers, revenue, or product maturity beyond the hackathon.

The author states:

• The tool was inspired by frustration with losing knowledge after research.
• It aims to make every research pack reusable as an agent skill.
• The system is designed for "agent-native" workflows and eval-first authoring.
• It supports multiple agents (Codex, Claude Code, Cursor, Copilot) via a shared standard.

Inference: The positioning is that of a tool for AI agent developers or researchers who want to turn knowledge into reusable, validated skills. It positions itself as an extension of the agent ecosystem.

Not evidenced: No claims about market fit, customer feedback, or competitive differentiation beyond self-description.

The description states:

• The tool is built for people doing deep research and wanting to reuse that knowledge in agent workflows.
• It supports 70+ agents via the Agent Skills standard.
• It targets developers or researchers who work with AI agents and want to author skills eval-first.

Inference: The ICP appears to be AI agent developers, researchers, or knowledge workers who are building or using agent-based systems.

Not evidenced: No evidence of actual customers, usage data, or segmentation beyond the creator’s own use case.

The description states:

• Skills can be published to skills.sh, a catalog site.
• The tool supports installation via npx and direct agent integration.
• There is no mention of pricing, monetization, or business model.

Inference: The tool appears to be open-source or free-to-use, with a possible catalog-based distribution model.

Not evidenced: No evidence of revenue streams, pricing, or commercialization plans.

The description states:

• SkillsKit is agent-native by design.
• It uses AI tools like GPT-5.6 and Codex in its pipeline.
• The system includes validators, CI checks, and eval-first authoring.
• It supports the Agent Skills standard (.agents/skills/).
• Research packs are produced with GPT-5.6 via researchkit.

Inference: The tool is built for integration into AI agent workflows and uses modern AI-assisted development practices.

Not evidenced: No evidence of scalability, performance metrics, or delivery pipeline maturity beyond the hackathon.

The description states:

• The author used it to publish a “whole shelf” of skills during the hackathon.
• The live catalog is at https://www.skills.sh/paldom.
• It was built in a hackathon window.
• No mention of users, adoption, or product traction beyond the creator.

Inference: The tool exists and has been used by its creator, but there is no evidence of external adoption or product maturity.

Not evidenced: No data on user base, customer engagement, or product usage beyond the author’s own use.

The description does not mention any competitors. It focuses on the author's own workflow and tooling.

Inference: The space may include tools for agent skill development, research automation, or AI agent orchestration, but no specific competitive landscape is described.

Not evidenced: No evidence of existing products, market players, or competitive positioning.

• No external adoption: The tool appears to be used only by the creator.
• Unproven market fit: There is no evidence that others are interested in or using this system.
• Highly self-contained: It’s built for a specific use case and may not scale beyond the author's workflow.
• Dependency on AI tools: Heavy reliance on GPT-5.6, Codex, and researchkit may limit portability or scalability.

Inference: The tool is in early development and lacks evidence of traction or commercial viability.

1. What external feedback have you received on the system?
2. Are there any users or adopters beyond yourself?
3. How do you plan to monetize or scale this product?
4. What are your plans for expanding beyond the hackathon prototype?
5. How do you intend to validate that skills work across different agents and environments?