Create, edit, and verify golem-powers skills following the standard SKILL.md structure. Provides templates for skill frontmatter, workflow files, adapter files, and eval fixtures. Includes validation checks for description length, trigger word coverage, and structural completeness. Use when creating a new skill from scratch, editing an existing skill's structure or content, adding workflows or adapters to a skill, or verifying a skill meets quality standards before deployment. Triggers on 'create skill', 'write skill', 'new skill', 'skill template', 'edit skill', 'skill structure'. NOT for: invoking existing skills (call them directly), superpowers skills (different structure and location), or skill-creator agent workflows. Use when this capability is needed.
Create, edit, and verify golem-powers skills following the standard SKILL.md structure. Provides templates for skill frontmatter, workflow files, adapter files, and eval fixtures. Includes validation checks for description length, trigger word coverage, and structural completeness. Use when creating a new skill from scratch, editing an existing skill's structure or content, adding workflows or adapters to a skill, or verifying a skill meets quality standards before deployment. Triggers on 'create skill', 'write skill', 'new skill', 'skill template', 'edit skill', 'skill structure'. NOT for: invoking existing skills (call them directly), superpowers skills (different structure and location), or skill-creator agent workflows. Use when this capability is needed.
Writing Golem-Powers Skills
Meta-skill for creating executable skills. Skills are tools that Claude can invoke and automatically execute.
Skill Structure
Every golem-powers skill MUST have this structure:
The execute: field is what makes a skill executable. When Claude loads a skill with execute: frontmatter, it MUST run that script IMMEDIATELY via Bash before any other action.
Dual Execution Patterns
Pattern A: Bash Script
Best for: Simple CLI wrappers, no dependencies, quick operations.
Use BASH_SOURCE pattern - All scripts MUST self-detect their location
Path Standard (CRITICAL)
All skill scripts MUST use this pattern for portability:
#!/usr/bin/env bashset -euo pipefail
# REQUIRED: Self-detect script location (works from any cwd)
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
SKILL_DIR="$(dirname "$SCRIPT_DIR")"# Now use absolute paths relative to skillsource"$SKILL_DIR/config.sh"# If needed
For project detection (when script needs prd-json/, package.json, etc.):
# Walk up from cwd to find project rootfind_project_root() {
localdir="$PWD"while [[ ! -d "$dir/prd-json" && "$dir" != "/" ]]; dodir="$(dirname "$dir")"doneif [[ -d "$dir/prd-json" ]]; thenecho"$dir"elseecho""fi
}
PROJECT_ROOT="$(find_project_root)"if [[ -z "$PROJECT_ROOT" ]]; thenecho"Error: Cannot find prd-json/ directory" >&2
exit 1
fi
Why this matters:
Skills are symlinked into a CLI-specific directory, but agents run from project directories: