ai-team

Accessibility patterns for AI agents working on UI. Covers interactive input labeling, button context, ARIA live regions, keyboard interaction, focus management, and competing UI state precedence. Used by Face (AC generation), Murdock (test queries), B.A. (markup), Lynch (review checklist), and Amy (probing).

Standard patterns for agent activity logging and completion signaling. Consult this skill when logging progress milestones with ateam activity createActivityEntry, or when signaling work completion with ateam agents-stop agentStop.

ateam CLI reference for all A(i)-Team API interactions. Consult this skill when running any ateam command, piping ateam output to python/jq/shell, reading item or mission data programmatically, or unsure which flag or subcommand to use.

Code quality, type safety, async, and API/database patterns for AI coding agents. Consolidated reference covering naming, function design, immutability, TypeScript type safety, async best practices, error handling, N+1 queries, pagination, and consistent API shapes — with TypeScript examples.

Reviews all unmerged code in the current branch for readability, security, correctness, types, and test quality. Produces findings by severity (Must Fix / Should Fix / Consider) with a parallelism estimate. Invoked standalone or as the review phase of /ai-team:review.

Defensive coding patterns for AI agents. Covers guard-before-operate, async error recovery, input validation consistency, URL encoding, resource cleanup, transient state clearing, functional state updates, network request hygiene (AbortController, path encoding, consistent error shapes), and never-weaken-safety-nets — with language-agnostic pseudocode examples.

Defines the one canonical mission-brief document every planning entry point (plan, review, bug-fix, bug-stomp) emits and sets as the mission's prdPath, so Frankie's DoD walk and Stockwell's PRD+diff review work unmodified regardless of which entry point started the mission.

Three-layer verification (static analysis, wiring trace, browser check) for catching integration gaps that unit tests miss. Loaded by Amy and Stockwell.

Instance pool claim/release protocol for pipeline agents (Murdock, B.A., Lynch, Amy). Consult this skill before agentStart (to claim your pool slot) and when calling agentStop (to understand how the CLI handles release and next-agent claiming automatically).

Complete-coverage reading protocol and structural map for mission PRD files. Consult this skill before reading a mission PRD — a single Read silently truncates near ~25k tokens, and analyzing a partially-read PRD produces confidently wrong conclusions.

Security patterns for AI coding agents. Covers injection prevention, secrets handling, API error responses, URL path encoding, and an OWASP quick reference — with TypeScript examples.

Native teams messaging protocol for pipeline agents. Consult this skill when sending or receiving START, ACK, ALERT, FYI, BLOCKED, or DONE messages, handling shutdown requests, or implementing peer-to-peer handoffs between agents.

Comprehensive test quality guardrails for AI testing agents. Banned anti-patterns with code examples, the litmus test, and positive guidance for writing tests that actually verify production behavior.

Work item structure, sizing, and acceptance criteria standards for the A(i)-Team pipeline. Consult this skill when creating work items, reviewing decomposition quality, sizing features, writing acceptance criteria, or determining item types.

Creates a new Product Requirements Document in prd/drafts/ by default. Runs a mini-discovery workshop to gather context, then produces a structured PRD scaled to the feature's complexity.