/aget-create-skill

Create a new AGET skill following the standard development process per SOP_SKILL_DEVELOPMENT.md.

Input

$ARGUMENTS

Mode Detection

Skill Creation Process

Step 1: Name Validation

Validate the skill name follows S-V-O convention:

Check for existing skill:

ls .claude/skills/ | grep -q "<name>"

If skill exists, STOP and report conflict.

Step 2: Determine Skill Number

Check specs/skills/INDEX.md for next available SKILL-XXX number.

Current registry:

• •SKILL-001 through SKILL-005: Existing
• •SKILL-006: Reserved for /aget-create-project
• •SKILL-007: This skill (/aget-create-skill)
• •SKILL-008+: Available

Step 3: Create Skill Specification

Create specs/skills/SKILL-XXX_<name>.md with:

1. •Purpose section
2. •Requirements (R-SKILL-XXX-001, etc.) in EARS format
3. •Test cases
4. •Traceability section

Step 4: Create SKILL.md

1. •Create directory: .claude/skills/<name>/
2. •Copy structure from templates/skill/SKILL.template.md
3. •Fill in required fields:

Required frontmatter:

---
name: <name>
description: <one-line description>
---

For skills with side effects (file writes, git operations):

disable-model-invocation: true

Instruction format (per L561 - Instruction Asymmetry):

• •Use PROHIBITIVE form: "NEVER", "SHALL NOT", "DO NOT"
• •Avoid affirmative form (~0% compliance)

Step 5: Update Index

Add entry to specs/skills/INDEX.md:

| SKILL-XXX | [<name>](SKILL-XXX_<name>.md) | ACTIVE | <primary finding> |

Step 6: Report

Output summary of created artifacts:

• •Skill spec location
• •SKILL.md location
• •Index update status

Output Format

## Skill Created: /aget-<name>

**Artifacts**:
- Spec: specs/skills/SKILL-XXX_<name>.md
- Implementation: .claude/skills/<name>/SKILL.md
- Index: specs/skills/INDEX.md (updated)

**Next Steps**:
1. Test skill invocation: `/aget-<name>`
2. Verify behavior matches spec
3. Create L-doc if significant learnings

Constraints

These are INVIOLABLE - you MUST NOT violate these constraints:

1. •NEVER create a skill with non-S-V-O name without documented exception
2. •NEVER overwrite an existing skill without explicit confirmation
3. •NEVER create SKILL.md without valid frontmatter (name, description)
4. •NEVER use affirmative constraints - use prohibitive form per L561
5. •DO check for existing skills before creating
6. •DO use templates/skill/SKILL.template.md as base
7. •DO include Traceability section in SKILL.md
8. •DO update specs/skills/INDEX.md after creation

Rationale

Per AGET theoretical grounding:

• •Stigmergy (Grasse): Skill artifacts enable cross-session coordination
• •Extended Mind (Clark/Chalmers): Templates and SOPs extend agent skill-creation capability
• •Evidence-First Design (L289): Specs before implementation
• •Instruction Asymmetry (L561): Prohibitive instructions for governance

This skill implements the "template-driven skill creation" pattern per SOP_SKILL_DEVELOPMENT.md.

Traceability