Lbh Skill Manager

When to use

Use this as the control panel for a user's installed legal skills. It gives an at-a-glance inventory — what's installed, which version, enabled or disabled, last reviewed, update available — and routes each item to the right lifecycle action.

Not for: discovering new skills (lbh-registry-browser), the security gate (lbh-skills-qa), or the individual mechanical actions themselves — this skill orchestrates and hands off to lbh-skill-installer, lbh-auto-updater, lbh-disable, and lbh-uninstall.

Method

1. Build the inventory from install records: name, version, enabled state, QA verdict date, source, and any applied conditions.
2. Flag health issues: skills past the staleness threshold, skills with an update available, skills whose last QA is older than a re-review interval (default 12 months).
3. Detect conflicts: two installed skills claiming the same capability + jurisdiction, or overlapping trigger phrases that would fire ambiguously.
4. Recommend one action per flagged item and name the skill that performs it — never perform destructive actions from here.
5. Surface the security posture: count of enabled skills with any elevated condition, so the user sees their exposure.
6. Keep it read-and-route: this skill reports and dispatches; it does not mutate installs directly.

Example

Inventory shows 6 legal skills. Two both claim nda-review / US, a conflict — you recommend disabling the lower-scored one via lbh-disable. One is 14 months since QA — you route it to re-review via lbh-skills-qa. One has v2.2 available — route to lbh-auto-updater. The user gets a single dashboard with three clear next actions.

Pitfalls

• Silently mutating installs from the manager. Keep it a router; destructive changes go through their dedicated, audited skills.
• Ignoring capability collisions, so two skills fire on the same request and produce contradictory legal output.
• Letting stale QA slide. A skill vetted a year ago against an older registry may now be risky; re-review on interval.
• Reporting enabled-count without exposure. Show which enabled skills carry elevated conditions.

Output format

# Legal Skills Inventory
INSTALLED (<n>):
- <name>@<ver> [enabled|disabled] QA:<date> update:<ver|none> cond:<...>
FLAGS:
- CONFLICT: <a> vs <b> on <capability/jurisdiction> -> lbh-disable
- STALE-QA: <name> (<months>) -> lbh-skills-qa
- UPDATE: <name> <cur>-><new> -> lbh-auto-updater
EXPOSURE: enabled-with-conditions=<n>

Reference

Skill-vetting checklist (management link): every installed entry must carry a QA date; entries past the re-review interval are surfaced for re-vetting, not left assumed-safe.

Security-review gate criteria: the manager tracks each enabled skill's applied conditions and re-review age; an enabled skill with a lapsed review is treated as a HIGH exposure flag.

Versioning / rollback: inventory shows installed vs available versions so updates and rollbacks target explicit coordinates; the manager routes, lbh-auto-updater executes with a rollback point.