Agent Skills
A collection of distributable skills for AI coding agents.
This repository is structured for compatibility with the skills CLI and follows the Agent Skills standard.
Available Skills
fast-algorithm-exploration-loop
Benchmark-driven workflow for converging on algorithms through small implementation spikes, fuzzing, visual diagnostics, and spec refinement.
Use when:
- exploring algorithms, numerical methods, geometry, graphs, optimization, or data structures
- multiple approaches are plausible and need empirical comparison
- success depends on measurable properties such as correctness, error, runtime, memory, or robustness
long-horizon-codex
Structure and run multi-hour Codex work with four control-plane documents: prompt.md, plans.md, implement.md, and documentation.md.
Use when:
- Scaffolding a repo for a long-running implementation effort
- Keeping multi-session work coherent across context compaction or handoff
- Creating durable execution plans and validation checkpoints
Inspired by Run long horizon tasks with Codex
Installation
Install the full collection from GitHub:
npx skills add phildionne/agent-skillsInstall a single skill from the repository:
npx skills add phildionne/agent-skills --skill long-horizon-codexRepository Structure
skills/<skill-name>/SKILL.mdcontains the skill definitionskills/<skill-name>/agents/openai.yamlcontains UI metadata when neededskills/<skill-name>/scripts/contains executable helpersskills/<skill-name>/references/contains load-on-demand documentationskills/<skill-name>/assets/contains templates and other output resourcesskills/<skill-name>.zipis the packaged distribution artifact