[Enhancement]: Add Agent Skills support to system prompt

[spachava753]

Enhancement Category

New MCP tool/server integration

Problem or Use Case

Currently, CPE's system prompt is static and doesn't support loading external "skills" that could extend the agent's capabilities in a modular way. Users who want to add domain-specific instructions, scripts, or references must manually modify prompts or rely on AGENTS.md files scattered throughout their projects.

The Agent Skills specification defines a standard format for packaging reusable agent capabilities. Supporting this format would allow users to:

• Share and reuse skills across projects
• Organize domain-specific knowledge into self-contained modules
• Progressively load instructions only when relevant (reducing context usage)

Proposed Solution

Add a template function that accepts multiple folder paths as arguments, where each path points to a skills directory. The function would:

1. Scan each provided directory for valid skill folders (containing SKILL.md)
2. Parse the YAML frontmatter to extract name and description fields
3. Inject skill metadata into the system prompt for agent awareness
4. Load full skill instructions on-demand when the agent activates a skill

Example usage in template:

{{ skills "./skills" "./shared-skills" "~/.cpe/skills" }}

The skill format follows the Agent Skills specification:

skill-name/
├── SKILL.md          # Required: frontmatter + instructions
├── scripts/          # Optional: executable code
├── references/       # Optional: additional docs
└── assets/           # Optional: static resources

SKILL.md frontmatter:

---
name: skill-name
description: What this skill does and when to use it.
license: MIT
metadata:
  author: example
  version: "1.0"
---

Rendering Format

Skills will be rendered in the system prompt using XML tags for clear structure:

<skills>
  <skill name="pdf-processing">
    <description>Extracts text and tables from PDF files, fills PDF forms, and merges multiple PDFs. Use when working with PDF documents.</description>
    <path>/Users/me/.cpe/skills/pdf-processing</path>
  </skill>
  <skill name="code-review">
    <description>Performs thorough code reviews with focus on security, performance, and maintainability.</description>
    <path>./skills/code-review</path>
  </skill>
</skills>

Alternatives Considered

• AGENTS.md files: Work well for project-specific context but aren't designed for reusable, shareable skills
• MCP servers: More powerful but heavier-weight; skills are better for instruction/knowledge-based extensions
• Custom system prompt editing: Requires users to manually manage prompt content

Impact Scope

Everyone using CPE

Additional Context

Reference specification: https://agentskills.io/specification.md

Key implementation considerations:

• Progressive disclosure: load metadata (~100 tokens) at startup, full instructions (<5000 tokens) on activation
• Validation: skill names must be lowercase alphanumeric with hyphens, match directory name
• File references in skills use relative paths from skill root