Skip to content
Chris & Mike edited this page Apr 9, 2026 · 14 revisions

Agent Skills

Memory Journal MCP Server includes shipped agent skills that can be auto-discovered by AI agents (like GitHub Copilot, Cursor, or Windsurf) to power advanced structured workflows.

GitHub Commander Skills

Introduced in v6.2.0, the GitHub Commander skill set provides a suite of robust workflows that coordinate GitHub integration tools, journal entries, and Code Mode evaluation. These skills ensure that all AI agent actions leave a proper audit trail in your project history.

Available Workflows

  • Issue Triage — Analyze incoming issues, assign labels, link to related journal entries, and add them to Kanban boards.
  • PR Review — Comprehensive PR review including code analysis, test coverage checks, and mapping to architectural decisions in the journal.
  • Roadmap Kickoff — Translate an approved epic implementation plan into an agent-actionable GitHub Kanban structure via Milestones and tracked Issues.
  • Sprint Milestones — Plan and track sprints by correlating GitHub Milestones with journal activity.
  • Security Audit — Review dependencies and data flow for vulnerabilities.
  • Code Quality Audit — Multi-step codebase audit checking against project standards and logging results.
  • Performance Audit — Check runtime performance and bundle sizes, documenting regressions.
  • Update Dependencies — Run a structured dependency update with validation gates and journal audit trail.

Discovery

Skills are exposed to the AI agent via the memory://skills resource.

  • Shipped Skills: Automatically discovered from the npm package.
  • Custom Skills: You can add your own skills by setting the SKILLS_DIR_PATH environment variable (or --skills-dir CLI flag) to point to a local directory containing your custom SKILL.md files.

Note: User-defined skills in SKILLS_DIR_PATH will override shipped skills with the same name.

Usage

When the agent reads memory://skills (e.g., at session start via memory://briefing), it becomes aware of the available workflows. You can then prompt the agent:

"Run the GitHub Commander PR Review skill on PR #42"

The agent will load the skill instructions on-demand and execute the structured workflow.

Clone this wiki locally