Agent Observability

This document describes the artifact capture layer that makes the desktop app inspectable by coding agents (Codex, Claude Code, Cursor) via existing WDIO/Appium/tauri-driver tooling.

Quick Start

bash app/scripts/e2e-agent-review.sh

Artifacts are at:

app/test/e2e/artifacts/<ISO-timestamp>-agent-review/
  01-welcome.png
  01-welcome.source.xml
  02-post-onboarding.png
  mock-requests-after-onboarding.json
  failure-<test>.png              # only on failure
  meta.json                       # run metadata + checkpoint index

Components

Component	Path	Role
Helper	app/test/e2e/helpers/artifacts.ts	captureCheckpoint, saveMockRequestLog
WDIO hook	app/test/wdio.conf.ts (afterTest)	Always dumps screenshot + source on any failing test
Spec	app/test/e2e/specs/agent-review.spec.ts	Welcome → onboarding → privacy panel
Wrapper script	app/scripts/e2e-agent-review.sh	Build + run + print artifact directory

Environment Overrides

Variable	Effect
E2E_ARTIFACT_DIR	Force specific run directory
E2E_ARTIFACT_ROOT	Parent directory (default: app/test/e2e/artifacts)
E2E_ARTIFACT_LABEL	Label (default: agent-review)

Using the Helper in New Specs

import { captureCheckpoint, saveMockRequestLog } from '../helpers/artifacts';
import { getRequestLog } from '../mock-server';

await captureCheckpoint('after-connect-click');
saveMockRequestLog('after-connect-click', getRequestLog());

Deliberately Out of Scope

• Visual baselines / image diff for each component state
• Screenshots on every click (too noisy)
• Live integrations (Gmail, Notion, Telegram); mock server only
• New test framework / reporter

Only expand to more flows after proving this loop is viable.

Next Steps

• E2E Testing - End-to-end testing
• Testing Strategy - Testing tiers