Skills

Reusable instruction packages that encode expertise into repeatable, consistent workflows.

The problem skills solve

Every team develops best practices for recurring tasks: how to write a commit message, how to review code for security issues, how to structure a pull request description. Without skills, this knowledge lives in people's heads or scattered wiki pages, and the AI assistant has to be re-taught every time.

Skills are structured instruction packages that capture this expertise once and make it available on demand. A skill defines the steps, criteria, and output format for a specific task, so PandaOS executes it the same way every time regardless of who triggers it. Think of them as saved expert workflows rather than saved prompts: they carry methodology, not just wording.

Skills vs. rules

Rules shape how PandaOS behaves passively across every interaction: coding conventions, naming patterns, tone preferences. They apply in the background without being invoked.

Skills are active: they run on demand to accomplish a specific task. A rule might say "always use conventional commit format," while a skill walks through the full commit workflow: analyze changes, draft a message, stage files, and create the commit. Rules set guardrails; skills execute playbooks.

The built-in catalog

PandaOS ships with a catalog of ready-made skills covering common development workflows: code review, git operations, test-driven development, frontend design evaluation, security audits, and more. Each catalog skill is maintained and updated with the platform, so improvements arrive automatically.

Installed skills are available across all your chats and projects immediately. You only need to install the ones relevant to your workflow since unused skills never interfere with normal conversations.

Running a skill

There are two ways to trigger a skill:

• Slash commands - Type / in the chat input to see available skills, then select one (for example /git-commit or /security-review). This is the explicit path: you know which workflow you want and invoke it directly.
• Automatic matching - Describe what you need naturally. When your request aligns with an installed skill, PandaOS recognizes the match and runs the skill's methodology without being told which one to use. For example, asking "review this PR for security issues" activates the security review skill automatically.

Skills also power Agents. When a specialized agent picks up a task, it invokes the appropriate skills behind the scenes, chaining multiple workflows together to complete complex work.

Creating custom skills

When the catalog does not cover a workflow specific to your team or project, you can create a custom skill. A custom skill defines the task objective, the step-by-step methodology, and any constraints or output formats.

For example, a team might create a "database migration review" skill that checks for backward compatibility, verifies rollback procedures, and flags missing indexes, specific to their schema conventions. Once saved, a custom skill behaves identically to a catalog skill: it appears in the slash command menu and can be matched automatically.

Because skills are stored with your project, everyone working in the same project shares the same workflows. This turns individual expertise into team capability: a senior engineer's code review checklist becomes a skill that every team member can invoke.