Electrified Cortex — Skills

Source library for AI-agent skills following the Agent Skills open standard. Compatible with Claude Code, VS Code (GitHub Copilot Chat), and any runtime that loads SKILL.md.

Every skill goes through a deliberate spec → write → audit cycle; the git history tells the full story.

Catalog

Each top-level directory is a skill. See skill.index.md for the structured index, or browse the directory listing.

Using these skills

Most users want the published plugin, not this source repo. Install instructions live in the skills-plugin distribution repo.

Contributing

Authoring workflow lives in skill-writing: spec first, derive an uncompressed source, compress, audit, stamp. The spec-writing, skill-auditing, and markdown-hygiene skills define the contracts contributors must satisfy.

Philosophy

Skills ship with companion .spec.md files that preserve design rationale and constraints in full natural language. The skill tells machines what; the spec tells humans and agents why.

Foundational journey

A small set of skills are load-bearing — everything else builds on them.

• compression — the first skill. Cuts token waste in agent contexts via three tiers (Lite / Full / Ultra). Inspired by JuliusBrussee/caveman (MIT); extended with safety gates and companion specs.
• spec-writing — the contract layer. Any file, feature, or system can have a behavioral spec; specs are the source code, runtime files the compiled program.
• skill-writing — the authoring pipeline. Spec first, derive uncompressed, compress to runtime, audit. Encodes the inline-vs- dispatch test and the rule that SKILL.md is never edited directly.
• spec-auditing + skill-auditing — the governance pair. Spec drift and skill quality are caught by zero-context dispatched audits before anything ships.
• dispatch — the architectural primitive. When to spawn a sub-agent, what model tier, what prompt shape; everything that delegates rests on this.

License

MIT — see LICENSE.