Project Description
I have an in-house AI agent that triggers k6 load test based on github push or github webhook events , but it still misses two critical situations: Identifying code updates and new feature additions and update the scripts or create new scripts accordingly , I need that gap closed.
I have around 2000 selenium and playwright scripts and 100+ loadrunner and k6 scripts , the agent should work on all those existing scripts
For development purposes i will share a test application repo , we can use this repo to evaluate the agent
The project is partially developed, it creates selenium, k6 and loadrunner scripts and deployed in aws bedrock agent core runtime . we need to add playwright and make the overall solution more robust to support complex flows like journey flows in the script .
Your task is to create a new agent so it watches a codebase , create k6 , loadrunner, selenium and playwright scripts for new functionality or update exiting scripts to add new functionality mentioned above .
let the implementation be Python only Once a change is spotted the agent must automatically adjust our existing automation and performance test scripts, regenerating or patching them so CI runs green without manual edits.
The current pipeline is circleci -based with k6 and loadrunner for performance checks; hooks are already in place, they just need smarter input from the agent.
Filter out any code or PII data :
GitHub Repo
↓
Code Scanner Agent
↓
PII/Secret Redaction
↓
Metadata Extractor
↓
LLM
↓
Generate k6 / Selenium / LoadRunner scripts
↓
Human review
↓
Commit generated scripts
Deliverables 1. New agent code with clear comments. 2. Detection logic covering the scenarios above, demonstrated with sample PRs. 3. Code should be run locally without pushing to github first for internal demo , then for production version it should be through github and cicd pipeline deployed in aws bedrock agent core runtime 4. A concise README explaining setup plus a one-command demo. 5. code walkthrough through anydesk
Acceptance will be based on a pull request against a staging branch, where your agent must detect a simulated library upgrade and a small feature file, then update the functional tests and performance scripts so all pipelines pass.