CI/CD Security
Chelate
Don't let your workflows work against you
Your pipelines are the easiest way in
Shrike scanned Threlte's GitHub Actions workflows and identified a critical vulnerability where an attacker could inject malicious code through workflow manipulation. The scanner flagged dangerous permission combinations and untrusted input flowing into pipeline execution contexts — exactly the kind of exploitable path that generic linters miss.
We reported the vulnerability privately to Threlte's maintainers with a detailed exploit path and recommended remediation. The team responded quickly, patched the workflow, and published a security advisory. No exploitation occurred. This is how responsible disclosure should work — find it, fix it, then tell the world.
With Titra installed, this vulnerability would have been blocked at runtime before any damage could occur. Titra monitors every action in your pipelines — git operations, credential access, package publishes — whether triggered by traditional CI/CD workflows or AI agents. You get complete visibility into what's running and the ability to block dangerous moves in real time.
/ The product
Free · Open source
find the risk
A CLI and GitHub Action that scans your workflows and tells you which pipelines an outsider can attack — ranked by Severity, Triggerability and Impact. No noise.
$ curl -fsSL https://chelate.tech/install.sh | bash $ shrike scan --github-org your-org
Enterprise
watch and stop it
Install the GitHub App, add one line to your workflow. Titra records every action your agent takes and blocks the dangerous ones — in real time.
/ How we secure them
01
Can someone with no access reach a secret or a write token — and how? That one question turns a pile of findings into the few paths that matter.
02
Titra sits in the running pipeline and sees every command, push and tool call. Finally, a record of what ran.
03
The dangerous move is often a normal-looking call to GitHub itself. We gate the agent's action — what firewalls can't see.
04
A model deciding what to block could be prompt-injected too. Our rules can't be talked out of a block.
05
Every install starts watch-only. See what your agents do, then turn on blocking when you trust it.
06
The scanner is free and open so the whole ecosystem can check itself. We charge for scale.
Backed by expertise — researchers, advisors & allies
Recognized for responsibly disclosing critical vulnerabilities in Microsoft's cloud & identity infrastructure
Critical vulnerabilities found and reported to Google, recognised in their security acknowledgements
All assessments follow OWASP Testing Guide v4.2 and Top 10 methodology across every engagement
Active CVE contributors — we find, report, and help patch real vulnerabilities in widely-used software
Top-ranked researchers across HackerOne, Bugcrowd and Intigriti with critical-severity findings
Trusted by maintainers of widely-used npm, PyPI, and Go packages to audit before major releases
Vulnerabilities found in AI coding agents — the same class of tool we now secure
Secured NASA systems and government infrastructure, with official acknowledgement for critical findings
/ How we secure the open source
Maintainers send us their packages ahead of a major release. We read the build, the dependency tree and the publish pipeline — not just the source.
Shrike is open source. Anyone can run it on their own org, read the rules, and check exactly what we would flag.
Findings go to the maintainer first, with a fix in hand. Write-ups come after a patch ships, never before.
npm · PyPI · Go modules · GitHub Actions · AI coding agents