Skip to content

skill-manager

Manage skill availability to control context and auto-activation behavior.

  • Area: Foundational
  • Category: utilities
  • Level: beginner
  • Input: conversation
  • Output: document
  • Setup: none
  • Roles:

Toggles vault skills by flipping disable-model-invocation in each skill's frontmatter, lists the active inventory by category and status, and bulk-enables or…

How you ask

  • “List all enabled skills and their categories”
  • “Disable the video skills to reduce context usage”
  • “Enable only the marketing-related skills”
  • “Show which skills are currently active”

Inputs

  • request — The inventory question to answer — which slice of the library the caller wants to see and at what grouping.
  • filter — Narrows the listing to one activation state. The choice decides whether the report can confirm a toggle took effect or only browse the library. (enabled | disabled | all)
  • pretty — Renders the listing as a wide table instead of the compact one-line-per-skill form.
  • request — The toggle to perform, stated as the one skill the caller means rather than a category that happens to contain it.
  • action — Direction of the toggle. Disabling removes the skill from auto-activation and from the context budget; enabling restores both. (enable | disable)
  • skill-name — Exact skill directory name under {skills.root}/. An alias or partial name is a pattern target and belongs on the pattern-or-category route.
  • request — The bulk change to perform, including the scope the caller believes it covers so the preview can be checked against that belief.
  • action — Direction of the toggle applied to every skill in the matched set. (enable | disable)
  • target — The match expression — all, a frontmatter category, an explicit category:<name>, a documented alias, or a shell-style folder pattern. Open-ended, because categories and aliases both move with the library.
  • dry-run — Prints the skills the target would match and changes nothing. Run it before any broad target so the matched set is confirmed rather than assumed.
  • request — Which indexes need refreshing and what changed upstream to make them stale.
  • indexes — The subset of derived indexes to rebuild. Absent, all three are regenerated, which is the safe default because they are read together.
  • shipping-only — Excludes private-tier playbooks from the workflow index, producing the index that matches a client-facing release rather than the internal library.

Outputs

  • Capability index — every enabled skill with its description, category, tier and value chains.
  • PLAYBOOKS.md
  • SKILLS.md
  • Stage-coverage map — which skills and playbooks cover each stage of each value chain, and which stages have none.
  • The listing as printed by list_skills.py, ending with the enabled and disabled counts.
  • The resolved match set, the skills whose activation state changed, the resulting enabled and disabled counts, and the restart reminder.
  • The skill touched, its before and after activation state, the resulting enabled and disabled counts, and the restart reminder.
  • VALUE-CHAINS.md
  • Workflow index — every routable playbook grouped by value chain.
  • Written inventory report — one row per skill with its activation state, category and tier — saved when the caller wants the listing as a file rather than on screen.
  • prefer a permanent directory)