---
name: uberdev-cmd-testers
description: "Use when the user wants to read-only adversarial QA audit squad. Spawns 8 agents (6 personas + 2 monitors) over 3 coordinated waves against an auto-detected target (web/api/native/all). Findings are evidence-anchored against a 10-invariant oracle library and auto-filed as GitHub issues via findings-to-issues. The squad never writes app code. Invokable explicitly as $uberdev-cmd-testers. Original description: Read-only adversarial QA audit squad. Spawns 8 agents (6 personas + 2 monitors) over 3 coordinated waves against an auto-detected target (web/api/native/all). Findings are evidence-anchored against a 10-invariant oracle library and auto-filed as GitHub issues via findings-to-issues. The squad never writes app code."
---

# Codex bridge — read first

This skill was ported from a Claude Code slash command (`/testers`). On Codex:

- **`$ARGUMENTS`** below = the user's free-text request (the words after the
  command name, or your whole task description if invoked implicitly).
- **`Task` tool** calls → use `spawn_agent`; collect results with `wait_agent`
  (see ~/.agents/skills/using-uberdev/references/codex-tools.md for the
  named-agent mapping).
- **`Skill` tool** invocations → skills load natively; just follow the named
  skill's instructions.
- **`Workflow` tool** (testers/uberscan/ubersimplify) → no Codex equivalent;
  follow the skill's `## No-Workflow fallback` section instead.
- **`MultiEdit`** → apply edits with your native file-edit tool.

Original argument hint: `[<target>] [--target=web|api|native|all] [--watch] [--rounds=N] [--max-issues=N] [--persona=name,...] [--no-issues] [--rps-cap=N]`

---



# Testers — Adversarial QA Audit Squad

Spawn an 8-agent read-only QA audit squad against the target in **$ARGUMENTS** (or auto-detect from CWD). Six distinct-persona testers (`panicked_grandma`, `power_user`, `adversarial_security`, `chaos_engineer`, `a11y_critic`, `mobile_thumb`) + two monitors (`monitor_primary`, `monitor_devils_advocate`) run three coordinated waves; monitors generate per-persona follow-up prompts between waves; findings without an `invariant_violated` field and an evidence anchor are dropped. Findings flow into `findings-to-issues` for durable GitHub-issue persistence.

**Usage:** `/uberdev:testers [<target>] [--target=...] [--watch] [--rounds=N] [--max-issues=N] [--persona=name,...] [--no-issues] [--rps-cap=N]`

- `<target>` — optional URL / OpenAPI path / binary path; auto-detect from CWD if omitted.
- `--target=<surface>` — explicit surface (`web | api | native | all`); overrides auto-detect.
- `--watch` — run the squad inline in this session via the No-Workflow fallback directive path (non-headless browsers, transcripts visible), instead of dispatching the background Workflow script.
- `--rounds=N` — wave count (default 3; minimum 1).
- `--max-issues=N` — cap for `findings-to-issues` (default 10).
- `--persona=name,...` — override default roster (drop a custom persona at `plugins/uberdev/agents/testers-<name>.md`).
- `--no-issues` — skip `findings-to-issues`; report only.
- `--rps-cap=N` — per-host RPS ceiling, [1, 1000], default 10. Enforced for curl traffic at lib/rate-limit-curl.sh; audited for Playwright/MCP traffic at Phase 5 (fail-the-run on breach).

**Read-only contract:** the squad has no `Edit` or general `Write` on app code; agent `allowed-tools` whitelists enforce it at the tool boundary.

Now invoke the `uberdev:testers-pipeline` skill — it owns the pipeline (parse + auto-detect, then the Workflow script `skills/testers-pipeline/workflow.js` runs the N-round waves, report synthesis, and findings persistence in the background; the thin preflight emits the Workflow args and states the post-Workflow politeBreach exit-1 mandate). A `## No-Workflow fallback` (the retained inline `--watch` directive path) covers platforms without the Workflow tool. The skill renders inline, so `$ARGUMENTS` remains in scope.