update-agent

Update selected agent definitions to the current split-source template while preserving role expertise.

4|1|Updated Jul 31, 2025
One-click install
npx skills add https://github.com/alvis/.agents --skill update-agent-alvis
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: update-agent
Source: https://github.com/alvis/.agents/tree/main/plugins/governance/skills/update-agent
Command: npx skills add https://github.com/alvis/.agents --skill update-agent-alvis

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agent definitions drift out of sync with evolving templates and harness requirements, and manual edits risk destroying role-specific expertise, trigger boundaries, or collaboration links. This Skill performs bounded, verified migrations of explicitly selected agent source sets without losing what makes each role useful. ## Core Features & Use Cases - Template Migration: Reconciles base.md and split frontmatter/{meta,claude,codex,grok}.json files with the current agent template while preserving voice, triggers, and expertise. - Behavior Corrections: Applies explicitly stated configuration or behavior changes, re-evaluating intelligence level, permissionMode, and isolation against classification criteria. - Multi-Harness Verification: Runs the deterministic stitch helper for Claude, Codex, and Grok targets and validates all JSON outputs before completion. - Use Case: After the agent template gains a new frontmatter convention, select plugins/coding/agents/reviewer and migrate it to the new revision while keeping its review expertise and trigger surface intact. ## Quick Start Update the agent at plugins/coding/agents/reviewer to the current template and report every changed field.

Frequently Asked Questions about update-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I migrate an agent definition to a new template revision?

Provide an exact agent path, name, or glob selector and optionally a --changes description. The skill snapshots the target's expertise and triggers, reconciles base.md and the four frontmatter JSON files with the current template, then verifies the result with the stitch helper for all three harnesses.

How do I update multiple agents at once?

Pass a glob selector or explicit --all flag for repository-wide work; an empty selector never means all. Independent targets can be delegated in bounded batches of one agent source set per subagent, within the declared child-spawn budget.

What is the difference between update-agent and create-agent?

update-agent handles bounded migrations and corrections to existing, explicitly selected agent definitions. create-agent owns genuinely new roles; any request that actually creates a new role is rejected and routed there.

Does updating an agent preserve its role-specific voice and triggers?

Yes. The workflow snapshots owned outcomes, positive and negative triggers, expertise, and working voice before editing, then integrates approved changes into existing prose rather than appending boilerplate. Completion requires no unrequested role, voice, permission, or ownership change.

Why does agent frontmatter validation fail after an update?

Common causes include malformed JSON, obsolete keys, a tools allowlist that should be omitted, or initialPrompt disagreeing with base.md. Parse all four JSON files with python -m json.tool and re-run the stitch helper for each harness to locate the failing check.