Ouroboros intent e2e
Skill secondorderai/ouroboros/.agents/skills/ouroboros-intent-e2e
Ouroboros is an Agent Harness, CLI and Desktop App with Self-Improvement layer
npx -y skills add secondorderai/ouroboros --skill ouroboros-intent-e2eAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Run Ouroboros desktop intent-based E2E test plans written in Markdown by driving the Electron app through Agent Browser over Chrome DevTools Protocol. Use when asked to execute, debug, or author natural-language desktop QA charters in test-plan/.
SKILL.md
2.2 KB, as published. Nobody here has run it
Ouroboros Intent E2E
Use this skill to execute Markdown test plans from test-plan/ against the real
Ouroboros Electron desktop app through Agent Browser CDP.
Workflow
- Read the assigned Markdown plan completely.
- Load version-matched Agent Browser guidance:
agent-browser skills get coreagent-browser skills get electron
- Connect to the launched app:
agent-browser connect {CDP_PORT}
- Capture initial state:
agent-browser snapshot -iagent-browser screenshot --annotate {OUTPUT_DIR}/initial.png
- Execute each intent step as a user would:
- Prefer accessible names and snapshot refs.
- Re-run
snapshot -iafter every click, navigation, modal, menu, or render-changing action. - Treat all
@eNrefs as stale after UI changes. - Use explicit waits for expected text or UI state.
- Before verdict, capture:
agent-browser consoleagent-browser errors- a final annotated screenshot
- Write
{OUTPUT_DIR}/report.mdand{OUTPUT_DIR}/result.json.
Rules
- Test only the running app. Do not inspect source files while executing the plan.
- Preserve evidence. Do not delete screenshots, videos, logs, or reports.
- If the UI blocks progress, capture the blocked state and return
INCONCLUSIVEunless a clear product failure is visible. - If an expected outcome is contradicted by visible UI state, return
FAIL. - If the plan passes but console/page errors appeared, include them in the report and choose
FAILwhen they affect user-facing behavior.
Result JSON
Write valid JSON with this shape:
{
"verdict": "PASS",
"summary": "Short outcome.",
"checks": [
{
"name": "Onboarding completed",
"status": "PASS",
"evidence": ["after-onboarding.png"]
}
],
"bugs": [],
"artifacts": ["initial.png", "final.png", "report.md"],
"consoleErrors": []
}
verdict and each check status must be PASS, FAIL, or INCONCLUSIVE.