companion-mode

Runs single-task Playwright verifications and produces evidence bundles with screenshots, video, trace, and HAR.

5|2|Updated May 19, 2026
One-click install
npx skills add https://github.com/civitas-cerebrum/achilles --skill companion-mode-civitas-cerebrum
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: companion-mode
Source: https://github.com/civitas-cerebrum/achilles/tree/main/skills/companion-mode
Command: npx skills add https://github.com/civitas-cerebrum/achilles --skill companion-mode-civitas-cerebrum

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @playwright/cli, @civitas-cerebrum/element-interactions.

What problem does it solve? QA engineers often need quick, ad-hoc proof that a specific flow works — for a ticket, a stakeholder demo, or a manual test checklist — without building a durable test suite or running a full coverage pipeline. This Skill runs one focused functional verification against a live app and delivers a complete, human-readable evidence bundle. ## Core Features & Use Cases - Evidence-first verification: Executes one task via Playwright and captures per-step screenshots, video, trace, HAR, and console logs into a self-contained bundle under tests/e2e/evidence/. - Structured verdicts: Reports PASSED, FAILED, or INCONCLUSIVE grounded in the user's verbatim pass criterion, with mandatory secret redaction of HAR and console artifacts. - Automation graduation: Offers to hand a passing run off to the durable test pipeline (Stage 3) or route failures to failure diagnosis. - Use Case: A QA engineer asks to verify the checkout flow on staging with proof; the Skill snapshots the pages, composes an instrumented spec, runs it, and delivers a bundle with a summary, video, and screenshots ready to attach to a ticket. ## Quick Start Ask the agent to verify a specific flow on a given URL with evidence, for example: verify the checkout flow on staging and capture screenshots, video, and a trace as proof.

Frequently Asked Questions about companion-mode

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

FAQPage Schema
How do I verify a user flow with screenshots and video evidence?

Provide a one-sentence task, the app URL, and a pass criterion. The skill discovers the page surface with playwright-cli, composes an instrumented Playwright spec, runs it with tracing and video on, and bundles per-step screenshots, video, HAR, and console logs.

What is the difference between companion mode and writing a durable Playwright test?

Companion mode produces an evidence bundle for a human to read, not a committed spec. Durable test authoring belongs to the achilles-protocol Stages 1-4; a passing companion run can optionally graduate into the suite afterward.

Does the evidence bundle get checked for leaked credentials?

Yes. Every captured network.har and console.log gets a mandatory redaction pass that strips credentials, API keys, PII shapes, and tokens in Authorization or Set-Cookie headers, with each redaction recorded in the bundle's summary.

What happens when a companion-mode verification run fails?

The bundle still ships with the failure evidence and a FAILED verdict. The skill then offers to hand off to failure-diagnosis to classify the cause as a test issue or an app bug before any automation is considered.

When should I not use companion mode for QA tasks?

Do not use it to grow a durable suite, expand coverage, hunt bugs adversarially, repair a rotted suite, or diagnose one failing test. Those belong to achilles-protocol, coverage-expansion, bug-discovery, test-repair, and failure-diagnosis respectively.