Claude Code skills¶
Seven task-shaped skills under _skills/ compose vstack's 34 patterns into real workflows.
| Skill | What it does | Patterns composed |
|---|---|---|
/vstack |
Meta entry — routes a free-form complaint to the right skill. | (meta only) |
/vstack-pick-pattern |
Interview-based pattern picker. | (meta only — surfaces all 34) |
/vstack-post-incident |
AAR → Lewin → 1-2 downstream patterns. | aar, lewin, dynamic downstream |
/vstack-audit-crew |
5-pattern crew health check. | lencioni, psych_safety, trust_triangle, process_gain_loss, bias_stack |
/vstack-bottleneck |
Load + structure diagnosis. | span_of_control, org_structure, social_loafing, superflocks |
/vstack-culture-check |
Surface-vs-assumption-layer culture audit. | schein_culture, robbins_culture, optional mcgregor |
/vstack-baseline |
Record + verify monitoring baselines. | All 34 |
Install¶
vstack-config install-skills # copies to ~/.claude/skills/vstack/
vstack-config install-skills --dry-run # preview first
vstack-config install-skills --force # overwrite existing
Restart your Claude Code client after install to pick up the skills.
Skill design — why task-shaped, not pattern-direct¶
Each pattern's MCP tool is already directly callable. Building 34 pattern-direct slash commands on top would duplicate that surface. The skill shape is different: each /vstack-* skill encodes the task ("we had an incident", "the crew is slowing down", "the culture feels off") and composes the right pattern bundle into one executive readout.
See the composition runbook for the canonical chains each skill orchestrates.