Envblex Skills is a local collection of AI-agent workflow skills maintained for envblex projects.

These skills define repeatable workflows for:

• repository bootstrap
• specification writing
• implementation
• review
• refinement
• explanation
• operations

Envblex Skills is not:

• a general-purpose skill registry
• an official OpenAI, Anthropic, Google, GitHub, or Cloudflare product
• a package manager
• a deployment system
• a replacement for Git permissions, branch protection, repository rules, or token restrictions

The skills are plain repository assets. They are intended to be copied, referenced, or installed into an agent runtime that understands skill-style instruction files.

Use skills.

pnpm dlx skills add envblex/agent-skills

skills/
  bootstrap-skill/
  explain-skill/
  implement-skill/
  ops-skill/
  refine-skill/
  review-skill/
  spec-skill/
templates/
  base/
  chrome-extension/
  go-cli/
  nodejs-api/
  nodejs-fullstack-webapp/

Each skill contains a canonical instruction file:

skills/<skill-name>/SKILL.md

Some skills may also contain agent-specific adapter files:

skills/<skill-name>/agents/

SKILL.md is the source of truth. Agent-specific adapter files must not weaken the rules defined in SKILL.md.

Initializes or normalizes a repository by creating project documents, task structure, scripts, and development conventions before implementation.

Typical output:

• SPEC.md
• TODO.md
• README.md
• AGENT.md

Turns rough project notes into a structured implementation specification.

Typical output:

• SPEC.md
• TODO.md
• README.md
• AGENT.md

Implements only the approved current scope defined by the project documents.

The implementation workflow must respect:

• PATCH_REQUEST.md, if present
• SPEC.md
• AGENT.md
• README.md
• TODO.md

Reviews repository changes against the approved scope and project rules.

Priority order:

1. scope violations
2. security issues
3. correctness issues
4. verification gaps
5. maintainability issues
6. documentation drift

Converts rough change notes into a focused patch request.

Typical output:

• PATCH_REQUEST.md

Explains code, specs, errors, repository state, or verification failures without changing files.

Defines operational rules for Git, verification, deployment boundaries, and completion gates.

Clone this repository locally:

git clone https://github.com/envblex/agent-skills.git
cd agent-skills

Install the skills by copying or symlinking the skills/ directory into the target agent workspace.

Use this when local changes in this repository should be reflected immediately.

mkdir -p ~/.local/share/envblex-skills
ln -s "$(pwd)/skills" ~/.local/share/envblex-skills/skills

Use this when a stable snapshot is preferable.

mkdir -p ~/.local/share/envblex-skills
cp -R skills ~/.local/share/envblex-skills/

Use this when the target agent expects one skill directory at a time.

mkdir -p ~/.local/share/envblex-skills/skills
cp -R skills/spec-skill ~/.local/share/envblex-skills/skills/
cp -R skills/implement-skill ~/.local/share/envblex-skills/skills/
cp -R skills/review-skill ~/.local/share/envblex-skills/skills/

Open the relevant SKILL.md file and provide it to the target agent as the active workflow instruction.

Examples:

skills/spec-skill/SKILL.md
skills/implement-skill/SKILL.md
skills/review-skill/SKILL.md
skills/ops-skill/SKILL.md

For repository bootstrapping, use:

skills/bootstrap-skill/SKILL.md

For patch refinement, use:

skills/refine-skill/SKILL.md

For explanation-only work, use:

skills/explain-skill/SKILL.md

Use the skills in this order for a new project:

1. bootstrap-skill
2. spec-skill
3. implement-skill
4. review-skill
5. ops-skill

For an existing project change, use:

1. refine-skill
2. implement-skill
3. review-skill
4. ops-skill

For explanation-only work, use:

1. explain-skill

Envblex Skills assumes strict separation between human authority and agent authority.

Agents may work on non-main branches.

Agents must not:

• push to main
• commit directly on main
• merge pull requests
• edit secrets
• create or rotate deploy tokens
• change deployment settings
• change repository permissions
• change branch protection or repository rules
• bypass checks
• deploy from a local machine

Production deployment should be controlled by GitHub Actions from main, not by local agent commands.

The documentation in this repository defines workflow rules. It does not replace actual GitHub permissions, token restrictions, branch protection, or repository rules.

This repository can define agent behavior, but it cannot enforce repository permissions by itself.

Actual enforcement must happen through:

• GitHub token scopes
• GitHub Secrets
• branch protection rules
• repository rulesets
• GitHub Actions permissions
• human review before merge

Do not give an agent a token that can perform actions the agent is not allowed to perform.

In particular, do not give an agent a token that can push to main if the intended policy is that agents cannot push to main.

The base project template expects these files:

SPEC.md
TODO.md
README.md
AGENT.md

Defines the current approved implementation scope.

Tracks deferred work. It is not automatically current scope.

Explains how to set up and use the generated project.

Defines repository-specific rules for future agents.

Before claiming a change is complete, run the relevant checks for the repository being modified.

For this repository, at minimum run:

git status -sb
git diff --check

If project-specific scripts are added later, document them here.

This repository is maintained as a local workflow toolkit for envblex projects.

It should remain small, explicit, and easy to audit.