jeromekneip

jeromekneip/ai-team-starterkit

Agent Claude Code

Design checklist (from the scout skill — apply before shipping a new agent).

jeromekneip/ai-team-starterkit

Agent Claude Code

Review agent — the judgment gate. Use to review a deliverable, draft, plan, or change for quality, correctness, and safety before it ships, and to verify that the matching procedure (skill) was applied. Deliberately read-only. Trigger phrases include "review this", "check before I send", "quality gate".

jeromekneip/ai-team-starterkit

Agent Claude Code

Research agent. Use for web research sweeps, fact-finding across multiple sources, market or tooling comparisons, and profiling professional roles or unfamiliar domains (including as the research step of the scout skill). Read-only — it researches and reports; it never writes workspace files.

jeromekneip/ai-team-starterkit

Agent Claude Code

Writing agent. Use for drafting and polishing documents, reports, summaries of provided material, communications, and README/user-facing text — especially multi-page deliverables worth isolating from the main conversation. Trigger phrases include "draft", "write up", "polish", "turn this into a document".

jeromekneip/ai-team-starterkit

Skill Claude CodeCodex

Add a new capability to the team — a new agent, a new procedure, or an upgrade to an existing agent. Use when the owner asks to "hire", "add an agent", "add a skill", or when a task lands that no existing agent or procedure fits. Also covers merging thin agents and retiring unused ones.

jeromekneip/ai-team-starterkit

Skill Claude CodeCodex

Document a repeatable process as a new skill or policy. Use when the owner says "make this repeatable", "document this process", "we should have a procedure for this", or when the same multi-step task has been done twice and will recur.

jeromekneip/ai-team-starterkit

Instructions file

Instructions for jeromekneip/ai-team-starterkit, covering nexus — operating model, how nexus works, the team, growing the team and hard rules.