Create Lazy

create-lazy — Declare a Lazy Skill

Generate a lazy-<name> placeholder that intercepts trigger keywords, logs each hit to /praxis/thinking/lazy/<name>.jsonl, and returns control to Claude Code for manual handling. See reference.md for the generated template, design rationale, and promotion workflow.

Args

/create-lazy <skill-name> "<short intent hint>"

• <skill-name> — kebab-case, no lazy- prefix (this skill adds it).
• <intent hint> — terse by design. Treated as intent, not final copy. Step 1.5 rewrites it into a trigger-rich description.

Template variables

Algorithm

Execute in order. Abort on failure.

Step 1 — Validate args

• Empty <name> or intent → print usage, stop.
• <name> not matching ^[a-z0-9]+(-[a-z0-9]+)*$ → ❌ <name> must be kebab-case.
• <name> starts with lazy- → ❌ Drop the "lazy-" prefix — this skill adds it.

Step 1.5 — Enrich the description

Rewrite the user's terse hint into a well-formed description. Required shape:

1. One-sentence capability summary — concrete verbs, specific domain.
2. Trigger list — Use when user says "<phrase1>", "<phrase2>", … — include the user's original phrase verbatim plus 4–8 plausible variations (synonyms, alternate verb forms, file extensions, domain jargon).
3. Scope clarifiers — formats, domains, contexts that disambiguate from adjacent skills.

Rules: stay truthful to intent (no invented capabilities); include obvious triggers (file extensions, format names); keep under 1024 chars; no meta-commentary about being a placeholder.

Example enrichment:

• User input: "convert xlsx"
• Enriched: Convert Excel spreadsheets (.xlsx, .xls) to markdown or CSV for LLM context. Use when user says "convert xlsx", "convert excel", "excel to markdown", "xlsx to csv", "read this spreadsheet", or mentions .xlsx / .xls files. Handles multi-sheet workbooks and preserves table structure.

Show enriched version to user for y / edit / abort approval before Step 2.

Step 2 — Locate source repo

AGENT_SKILLS=/home/mat/dev/agent-skills
test -d "$AGENT_SKILLS" || { echo "❌ agent-skills source not found"; exit 1; }

Step 3 — Collision checks

# A. Already-lazy
[ -d "$AGENT_SKILLS/lazy/skills/lazy-<name>" ] && { echo "❌ lazy-<name> already declared."; exit 1; }

# B. Future-shadow — real <name> exists in any plugin
EXISTING=$(find "$AGENT_SKILLS" -maxdepth 4 -type d -path "*/skills/<name>" 2>/dev/null)
[ -n "$EXISTING" ] && { echo "❌ Real skill <name> exists at $EXISTING — would collide on promotion."; exit 1; }

Step 4 — Write the generated SKILL.md

Use the Write tool (creates missing parents) to write $AGENT_SKILLS/lazy/skills/lazy-<name>/SKILL.md. The file body is the template in reference.md § "Generated template", with <name> and <enriched-description> substituted. No leftover placeholders.

Step 5 — Bump the lazy plugin version

Delegate to /edit-plugin — never hand-edit version files. Call as a Skill invocation:

/edit-plugin lazy patch "add lazy-<name> placeholder skill"

patch, not minor. Rationale in reference.md § "Design rationale": lazy skills are experimental stubs, not real capability additions, so minor bumps inflate the version. /edit-plugin will pause with a review summary before finalizing — that pause is expected. Surface any error and stop; no manual recovery.

Step 6 — Confirm

✅ lazy-<name> declared at lazy/skills/lazy-<name>/SKILL.md
📊 First trigger will log to /praxis/thinking/lazy/<name>.jsonl
📦 lazy plugin bumped to <new-version>
🔌 Reload plugins (or restart CC) so the new skill is discovered.