[CNCF LFX Proposal] Podman Container Tools: Agentic CI Flake Categorization and Analysis

### CNCF Project

Podman Container Tools

### Term

2026 Term 3 (Sep-Nov)

### Program Name

Agentic CI Flake Categorization and Analysis

### Program Description

## Description
Continuous Integration (CI) pipelines are the backbone of our development, but "flaky" tests—tests that exhibit both false positive and false negative outcomes randomly—severely degrade developer velocity and erode trust in the CI system.
Currently, our project relies on GitHub Actions for its CI/CD processes. Manually digging through extensive GitHub Actions logs to identify, categorize, and troubleshoot these flaky failures is a massive, time-consuming burden for maintainers.
This internship aims to solve this by building an intelligent, automated toolchain that monitors our GitHub Actions workflows for flaky behavior and leverages an agentic AI workflow to handle the heavy lifting of analysis. The intern will build a system that automatically extracts failing logs, uses AI agents to reason about the failure (differentiating between infrastructure blips, race conditions, network timeouts, etc.), and surfaces actionable mitigation strategies directly to the maintainers.

## Expected Outcomes
Data Ingestion Pipeline: A mechanism to automatically fetch, filter, and parse flaky CI run data and logs directly from the GitHub Actions API.
Agentic Analysis Engine: An integration with an AI/LLM framework designed to read failure logs, categorize the root cause of the flake, and generate a plain-English analysis.
Mitigation & Reporting: A reporting layer that takes the agent's findings and seamlessly integrates them into the developer workflow (e.g., auto-generating GitHub Issues, compiling weekly flake reports, or posting PR comments with suggested fixes).
Documentation: Comprehensive documentation covering the architecture of the tool, how to deploy it, and how maintainers can tweak the agent's prompts and behaviors.

# Recommended Skills
Proficiency in Python or Go (for scripting and API interactions).
Familiarity with GitHub Actions, CI/CD concepts, and log analysis.
Interest or experience in Generative AI, LLMs, and agentic workflows (prompt engineering, AI tool calling).
Experience with the GitHub API is a strong plus.
Familiarity with Local AI is a plus

### Technologies

AI, CI/CD, Github Actions, Go, Python

### Skills same as Technologies?

- [x] Yes, the required skills are the same as the technologies listed above.

### Required/Desirable Skills

_No response_

### Mentors

Paul Holzinger | @Luap99 | pholzing@redhat.com | luap99
Tim Zhou | @timcoding1988 | tizhou@redhat.com | timcoding1988
Mohan Boddu | @mohanboddu | mboddu@redhat.com | bhujji

### Upstream Issue URL

https://github.com/podman-container-tools/podman/issues/29265

### Application Prerequisites

- [x] Resume
- [x] Cover Letter
- [ ] School Enrollment Verification
- [ ] Participation Permission from school or employer
- [ ] Coding Challenge
- [ ] Custom Prerequisite (fill in details below)

### Coding Challenge URL

_No response_

### Custom Prerequisite Name

_No response_

### Custom Prerequisite Description

_No response_

### Custom Prerequisite — File Upload

- [ ] Yes — completion of this task requires the mentee to submit a file.

---
**LFX program:** [CNCF - Podman Container Tools: Agentic CI Flake Categorization and Analysis (2026 Term 3)](https://mentorship.lfx.linuxfoundation.org/project/050e89d9-aec2-47ad-9113-3ba41a639d55)

Start by reading the upstream issue #29265 and reviewing the stated GitHub Actions API, log-analysis, agentic workflow, and reporting requirements. Done means delivering an automated pipeline for flaky-run ingestion, AI-based categorization, maintainer-facing mitigation reports, and deployment and prompt-tuning documentation.

Written by the indexing model from the issue text.