Editorial Methodology

AI Builder Library helps readers discover and compare prompts, Agent Skills, MCP servers, and agent frameworks. This page explains how an entry is selected, what we verify, what our labels mean, and how corrections are handled.

Selection criteria

We prioritize resources that solve a recognizable workflow, have a public source or authoritative documentation, identify a maintainer, and provide enough information for a reader to evaluate fit. For developer tools, we prefer first-party repositories maintained by the product or protocol owner. We may include a notable community project when its ownership is clear and its value cannot be represented by an official alternative.

Inclusion is editorial, not paid placement. A listing does not mean the resource is secure, suitable for every environment, or endorsed by AI Builder Library.

Source verification

Before publishing a developer resource, we open the linked source and compare the entry against the maintainer’s repository or documentation. We record the resource type, maintainer, stated license, supported runtime, compatibility claims, installation command, and operational status when those details are available.

The Official source label means the linked page appears to be controlled by the named maintainer or product owner. It is a provenance label—not a security certification. The Last reviewed date records when the directory entry was compared with its cited source. It is not necessarily the software’s release date.

Practical and safety review

Each resource page explains what the project does, where it is a good fit, and what a builder should check before installation. For Agent Skills, that includes triggers, scripts, referenced files, network access, and license boundaries. For MCP servers, it includes transport, credentials, exposed tools, directory scope, and remote-service permissions. For agent frameworks, it includes runtime, state, tracing, persistence, deployment, and the operational cost of additional orchestration.

We recommend least-privilege access, isolated testing, explicit approval for irreversible actions, and validation of model output. These notes are general risk controls and do not replace a security review for the reader’s environment.

Prompt review

Prompt pages are evaluated for a clear task, sufficient context, useful constraints, expected output shape, and a practical way to review the result. A prompt is not presented as a guarantee of accuracy. Readers should verify claims, protect confidential information, and add domain-specific checks for high-impact decisions.

Writing and AI assistance

Automation and AI tools may assist with drafting, formatting, categorization, and consistency checks. Material factual claims about listed software should remain grounded in the cited first-party source. Final publication decisions, source selection, page structure, and risk notes are governed by this methodology. We do not invent hands-on test results or claim a feature was verified in production when it was only documented by the maintainer.

Updates and corrections

Software and documentation change quickly. We review entries periodically and when we receive a credible correction. A material update may change the review date, compatibility, status, installation guidance, or safety note. We may remove an entry when its source disappears, ownership becomes unclear, the project is archived without a useful successor path, or the listing can no longer be represented accurately.

To report an error, email [email protected] with the page URL, the statement to correct, and a first-party source supporting the change. Maintainers may also request an ownership clarification or removal. We evaluate the evidence and update the page when warranted; we do not accept payment to suppress a factual correction.

Scope and limitations

The directory is not exhaustive. Repository activity, stars, popularity, and an open-source license do not prove security or quality. Compatibility can depend on version, host configuration, account permissions, geography, and provider policy. Always inspect the current source and official documentation before production use.