uipath-test-cloud-agent
MukundaKatta/uipath-test-cloud-agent
Agentic test-management layer for UiPath Test Cloud, built for UiPath AgentHack 2026 Track 3. Uses Claude Code + @mukundakatta agent-stack libs.
Instalação
npx convez-ai MukundaKatta/uipath-test-cloud-agent1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
uipath-test-cloud-agent
An agentic test-management layer for UiPath Test Cloud that uses Claude Code to generate test scenarios from requirements, executes them on Test Cloud, and uses my published @mukundakatta agent-stack libraries to flag fragile or drifted tests before they reach a release.
Built for: UiPath AgentHack 2026 — Track 3: UiPath Test Cloud (deadline Jun 29, 2026, 11:45 PM PDT).
Status: Proposal scaffold. UiPath Labs access requested 2026-05-18; estimated arrival 2026-05-21. Real integration code lands once the sandbox is live.
The brief (Track 3, quoted from the rules)
Create agents that use UiPath Test Cloud to reimagine how software testing is designed, automated, executed, and managed across modern enterprise environments. Your goal is to show how agentic software testing can improve quality across AI-driven automations, enterprise applications, and connected business workflows.
Track 3 explicitly calls out four agent capabilities:
- Evaluate requirements and turn them into meaningful test scenarios.
- Identify fragile or outdated tests before they slow down a release.
- Recommend fixes when automation breaks.
- Orchestrate the right tests at the right time based on risk, coverage, and change impact.
This project maps each capability to a published, tested library plus a thin Test-Cloud-orchestration layer.
Architecture
+--------------------+ +-------------------------+
| Requirements doc | | Production test runs |
| (PDF, Jira, etc.) | | on UiPath Test Cloud |
+---------+----------+ +------------+------------+
| |
| extract | snapshots + telemetry
v v
+---------+--------------------------------+------------+
| |
| Claude Code (or any coding agent |
| via UiPath for Coding Agents CLI) |
| |
+---+-------------+---------------+-----------------+---+
| | | |
| generate | validate | snapshot | score
v v v v
+---+----+ +---+----+ +-----+----+ +-------+--+
| TS | | arg | | trace | | drift |
| (gen) | | check | | diff | | / aiaud. |
| ai_aud | |agentvet| |agentsnap | |driftvane |
+--------+ +--------+ +----------+ +----------+
| | | |
+-------------+-------+-------+-----------------+
|
v
+---------+---------+
| UiPath Test Cloud |
| (Maestro / API |
| orchestration) |
+-------------------+
Capability ↔ library mapping
| Track 3 capability | Library | What it does today | Test Cloud role |
|---|---|---|---|
| 1. Generate test scenarios from requirements | aiaudit (repo) | Scores diffs for AI-generated content | Reverse it: ask Claude Code to generate test cases, then aiaudit scores them for quality/risk before they land. |
| 2. Flag fragile or outdated tests | agentsnap (npm) | Records and replays agent-trace snapshots | Snapshot every Test Cloud run; diff against baseline; flag tests whose behavior drifts even when assertions still pass. |
| 3. Recommend fixes when automation breaks | agentvet (npm) + agentcast (npm) | Validate tool args with LLM-friendly retry hints; structured-output enforcer | When a test fails, ask the coding agent for a fix and route through agentvet to validate, agentcast to constrain. |
| 4. Orchestrate by risk, coverage, change impact | driftvane (PyPI) | Composable RAG/agent drift detectors | Score "is this a real test failure or just upstream drift?" so the orchestrator can skip noise and run the right tests first. |
Bonus: coding-agent integration
The UiPath AgentHack rules give bonus points for solutions that use coding agents (Claude Code, Codex, Cursor, Gemini CLI) through UiPath for Coding Agents. This project is built around that pattern from the start:
- Test generation: Claude Code with an MCP server wrapping
aiauditfor scoring. - Failure diagnosis: Claude Code running with the
agent-safety-mcpsafety layer. - Orchestration: UiPath Maestro / Test Cloud is the runtime; coding agents are the brain.
Demo scenarios
Once Labs access lands and the agent is wired, the demo video will show:
- Drop a 3-page requirements document into the agent. It produces a UiPath Test Cloud test suite with
aiaudit-scored test cases ranked by risk. - Re-run yesterday's test suite.
agentsnapflags two tests whose telemetry drifted even though they passed. The agent explains why. - Force a failure. The agent proposes a fix via Claude Code,
agentvetvalidates the suggested patch, and the orchestrator re-runs only the affected tests.
Repo layout
src/ # UiPath Coded Agent source (Python SDK)
generator/ # Requirements -> test cases
scorer/ # aiaudit + agentcast wrappers
drift/ # agentsnap + driftvane wrappers
orchestrator/ # Test Cloud / Maestro glue
submissions/
uipath-agenthack.md # Devpost submission text (draft)
docs/
video-script.md # 3-min demo walkthrough script
tests/
test_imports.py # Smoke tests, no Labs required
Build plan + timeline
- 2026-05-18: scaffold proposal + Devpost registration (done)
- 2026-05-21 (est.): Labs access lands; spin up Test Cloud sandbox
- 2026-05-21 → 05-25: learn UiPath Test Cloud + Coded Agents SDK
- 2026-05-26 → 06-08: build generator + scorer + drift + orchestrator
- 2026-06-09 → 06-15: end-to-end demo on a real Test Cloud project
- 2026-06-16 → 06-22: record demo video, polish submission, write feedback for Best Product Feedback bonus
- 2026-06-23 → 06-29: buffer, final fixes, Devpost upload
- 2026-06-29 11:45 PM PDT: submission deadline
License
Apache 2.0