Skills Plugin

The Skills plugin implements the Agent Skills open standard: reusable capabilities the agent can pick up, invoked with a slash command such as /review or /write-tests.

Using skills

Type a slash command that matches a skill name at the start of your message:

/review
/review src/api.py
/write-tests MyClass

The skill's instructions are added as context, and any text after the command is passed through as your request. A card in the conversation shows which skill ran.

The agent is also told which skills exist, so it can follow one when your task matches even if you do not type the slash command.

Disable individual skills in Settings → Agent Skills → Disabled Skills.

Skill discovery

Skills are loaded from four places. Later sources override earlier ones with the same name:

System — skills bundled with Codumentor (when builtin_enabled is true)
Organization — directories listed in skill_dirs
Project — .codumentor/skills/ inside each indexed repository (when project_skills_enabled is true)
User — ~/.codumentor/skills/ (or user_skills_dir)

Writing a skill

A skill is a directory whose name matches the skill name, containing a SKILL.md file:

review/
├── SKILL.md        # Required: frontmatter + instructions
├── scripts/        # Optional: helper scripts
└── references/     # Optional: extra documents

SKILL.md example:

---
name: review
description: Perform a thorough code review focusing on security and correctness
license: MIT
---

You are an expert code reviewer. Analyze the provided code for:
1. Security vulnerabilities (injection, auth, secrets)
2. Logic errors and edge cases
3. Performance issues
4. Code style and maintainability

Provide a structured review with severity ratings.

The name must be 1–64 characters, lowercase letters, numbers, and hyphens, starting with a letter, and must match the directory name.

Configuration

plugins:
  - module: codumentor.plugins.skills
    class: SkillsPlugin
    args:
      enabled: true
      builtin_enabled: true
      project_skills_enabled: true
      user_skills_dir: ~/.codumentor/skills
      max_active_skills: 5
      skill_dirs:
        - skills/                     # Relative to the config file directory
        - /opt/shared-skills
      trust_policy:
        allow_scripts: true
        allowed_sources_for_scripts:
          - system
          - organization
        require_approval_for:
          - project

Parameter	Type	Default	Description
`enabled`	bool	`true`	Enable or disable the plugin.
`builtin_enabled`	bool	`true`	Load skills bundled with Codumentor.
`project_skills_enabled`	bool	`true`	Load skills from `.codumentor/skills/` in each indexed repo.
`user_skills_dir`	string	`~/.codumentor/skills`	Personal skills directory.
`max_active_skills`	int	`5`	Maximum number of skills injected into a single prompt.
`skill_dirs`	list	`[]`	Extra organization-level skill directories. Relative paths resolve against the config file directory.
`trust_policy.allow_scripts`	bool	`true`	Allow skill helper scripts at all. When false, the agent is told not to run scripts from any skill.
`trust_policy.allowed_sources_for_scripts`	list	`["system", "organization"]`	Skill sources whose `scripts/` directories the agent may use.
`trust_policy.require_approval_for`	list	`["project"]`	Skill sources the trust policy marks as needing approval.

Notes

Skills and Slash Commands can both listen for /… at the start of a message. Slash commands run first: if you have a personal command with the same name as a skill, the command expansion wins.