# GitInject: Real-World Prompt Injection Attacks in AI-Powered CI/CD Pipelines

DevFeed: [GitInject: Real-World Prompt Injection Attacks in AI-Powered CI/CD Pipelines](<https://devfeed.tech/articles/gitinject-real-world-prompt-injection-attacks-in-ai-powered-ci-cd-pipelines-113995.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.09935>)

Author: Jafar Isbarov, Umid Suleymanov, Ilia Shumailov, Murat Kantarcioglu

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [prompt injection](<https://devfeed.tech/topics/prompt-injection.md>), [ai security](<https://devfeed.tech/topics/ai-security.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [software supply-chain attack](<https://devfeed.tech/topics/software-supply-chain-attack.md>), [GitHub Actions](<https://devfeed.tech/topics/github-actions.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-powered](<https://devfeed.tech/tags/ai-powered.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [attacks](<https://devfeed.tech/tags/attacks.md>), [availability](<https://devfeed.tech/tags/availability.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [ci-cd](<https://devfeed.tech/tags/ci-cd.md>), [ci-cd-workflows](<https://devfeed.tech/tags/ci-cd-workflows.md>), [configuration](<https://devfeed.tech/tags/configuration.md>), [continuous-delivery](<https://devfeed.tech/tags/continuous-delivery.md>), [continuous-integration](<https://devfeed.tech/tags/continuous-integration.md>), [contribution](<https://devfeed.tech/tags/contribution.md>), [contributions](<https://devfeed.tech/tags/contributions.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [pipelines](<https://devfeed.tech/tags/pipelines.md>), [prompt-injection](<https://devfeed.tech/tags/prompt-injection.md>), [real-world](<https://devfeed.tech/tags/real-world.md>)

## AI overview

GitInject evaluates prompt injection in live, fully automated CI/CD workflows, where unmerged contributions can trigger AI agents with repository tools and credentials. The framework provisions ephemeral repositories and runs actual GitHub Actions workflows, capturing sandbox, credential, and permission behavior that simulated tool-call benchmarks may miss. Across eleven attack scenarios, five workflows, and four AI providers, each provider was susceptible to at least one high-severity attack in the tested configurations. The central weakness was structural: attacker-controlled configuration files could be loaded as operator-level instructions, potentially exposing credentials or corrupting review judgments. Two configuration-file attacks also replicated on GitLab and Hugging Face. The authors assessed workflow-level countermeasures and their coverage and utility tradeoffs; the supplied abstract does not specify which measures or tradeoffs.

## Source excerpt

arXiv:2606.09935v2 Announce Type: replace Abstract: AI agents are increasingly embedded in continuous integration and continuous delivery (CI/CD) workflows to autonomously review pull requests, triage issues, and maintain codebases. We study fully automated configurations in which external contributions invoke agents without prior human approval. Because these agents ingest untrusted content while operating with repository credentials and tools, prompt injection can have software supply chain consequences. We present GitInject, an open-source framework for evaluating prompt injection in live GitHub Actions workflows. Unlike benchmarks that simulate tool calls, GitInject provisions ephemeral repositories and executes actual workflows, exposing platform semantics such as sandbox constraints, credential handling, and permission boundaries. We use GitInject to evaluate eleven named attack scenarios across five workflows from four AI providers, covering configuration-file injection, credential exfiltration, judgment manipulation, and availability. Every provider is susceptible to at least one high-severity attack in the tested configurations. The most consequential vulnerabilities are structural: attacker-controlled configuration files from an unmerged contribution are loaded as operator-level instructions, allowing the agent--infrastructure interaction to expose credentials or corrupt review judgments. Two configuration-file attacks replicate on GitLab and Hugging Face as well as GitHub. We evaluate workflow-level countermeasures and show their coverage and utility tradeoffs. We release GitInject and all attack scenarios.