Research

These sources and recorded checks support the skills, installation methods, and documentation. Distinguish vendor recommendations and observed results from editorial inference.

The repository-wide skill and instruction system audit lives in Harness Atlas because it reviews notes-search agent infrastructure rather than one portable public skill.

Read the merged Andrej Explain research for Karpathy’s original post, related writing, learning design, skill-authoring practices, and evidence limits.

The current documentation is grounded in:

• the official OpenAI Prompting guidance for GPT-5.6 Sol;
• the official Anthropic Prompting Claude Fable 5;
• Zach Miles’s site, which inspired the name and original spark for Zach Prompting;
• the canonical HTMA Measure skill source;
• the Open Skills CLI repository and discovery format;
• the Skills CLI reference;
• the Claude Code plugin marketplace specification; and
• clean-room installation checks for the Open Skills CLI, Codex, and Claude Code completed on 2026-07-11 and 2026-07-15.

Future source-backed skill research can be published here as dated pages. Each memo becomes canonical at its deployed documentation path and appears as a separate Research sidebar entry.

• Concise, useful environment examples, user preferences and primary-source comparisons for Sam’s Env.

• Task-focused docs and a checked skill catalog, Stripe, Cloudflare, and Google practices applied to creator docs and collection maintenance.