Agent-Skills.md

Agent Skills: Validate Counts

Validates hook, skill, and agent counts are consistent across CLAUDE.md, hooks.json, manifests, and source directories. Use when counts may be stale after adding or removing components, before releases, or when CLAUDE.md Project Overview looks wrong.

UncategorizedID: yonatangross/orchestkit/validate-counts

Repository

yonatangross/orchestkit
yonatangrossLicense: MIT
13314

Install this agent skill to your local

pnpm dlx add-skill https://github.com/yonatangross/orchestkit/tree/HEAD/src/skills/validate-counts

Skill Files

Browse the full folder contents for validate-counts.

Download Skill

Loading file tree…

src/skills/validate-counts/SKILL.md

Skill Metadata

Name
validate-counts
Description
Validates hook, skill, and agent counts are consistent across CLAUDE.md, hooks.json, manifests, and source directories. Use when counts may be stale after adding or removing components, before releases, or when CLAUDE.md Project Overview looks wrong.

Validate Counts

Checks that hook, skill, and agent counts are consistent across all authoritative sources in OrchestKit. Outputs a comparison table and flags drift with precise file references.

Quick Start

# Full validation: counts src/ vs CLAUDE.md and manifests (run from repo root)
bash src/skills/validate-counts/scripts/validate-counts.sh

# Just get raw counts from src/
bash src/skills/validate-counts/scripts/count-all.sh

Rules

| Category | Rule | Impact | Key Pattern | |----------|------|--------|-------------| | Count Sources | rules/sources-authoritative.md (load ${CLAUDE_SKILL_DIR}/rules/sources-authoritative.md) | HIGH | Filesystem is authoritative; derived sources must match | | Drift Detection | rules/drift-reporting.md (load ${CLAUDE_SKILL_DIR}/rules/drift-reporting.md) | HIGH | Comparison table + flag with file:field references |

Total: 2 rules across 2 categories

Workflow

  1. Run scripts/validate-counts.sh for full validation (counts + drift comparison), or scripts/count-all.sh for raw counts only
  2. Read CLAUDE.md — extract counts from Project Overview and Version section
  3. Read manifests/ork.json — check skill/agent/hook array lengths
  4. Build the comparison table (see rules/drift-reporting.md for format)
  5. Flag any mismatches with file + field references; otherwise output "All counts consistent."

References

Load on demand with Read("${CLAUDE_SKILL_DIR}/references/<file>"):

| File | Content | |------|---------| | count-locations.md | Where every count lives and why drift happens |

Related Skills

  • release-checklist — Uses validate-counts as step 5 of the release gate
  • doctor — Broader health check that includes count validation
  • audit-skills — Quality audit for skill structure and completeness

Common Mistakes

  1. Counting from plugins/ instead of src/ — plugins/ may be empty after an interrupted build
  2. Comparing against deleted alias manifests — only manifests/ork.json exists in v7
  3. Forgetting the hook breakdown: global + agent-scoped + skill-scoped must sum to total