- Home
- Skills
- DevOps & Cloud
- Azure Pipelines Safety Check
More screenshots
Works with the AI Agents you already use
Azure Pipelines Safety Check
This skill finds those and the other Azure Pipelines mistakes below, and tells you how to fix each one.
$5
Azure Pipelines Safety Check
Example session with this skill installed
Check this project with azure-pipelines-safety-check.
- Read your context and instructions
- Compiled the azure pipelines safety
$ python3 scripts/check_azure_pipelines.py examples/risky
BLOCK azure-pipelines.yml:31 A01 SECRET_IN_PIPELINE
value: <redacted> [SONAR_TOKEN]
Why: A password, token or key is written into the pipeline YAML (a variable, an env entry, a task input, a parameter default, a URL, a script line, or a known token format such as ghp_ or npm_). Everyone who can read the repository can read it, forks and old commits keep it, and Microsoft says not to set secret variables in YAML.
Safer: Rotate the secret. Keep it as a secret pipeline variable, in a variable group (optionally linked to Azure Key Vault) or fetch it with the AzureKeyVault task, and map it into the step with env: (MY_TOKEN: $(myToken)).
BLOCK azure-pipelines.yml:45 A02 UNTRUSTED_TEXT_IN_SCRIPT
BLOCK azure-pipelines.yml:46 A03 INSECURE_DOWNLOAD_EXECUTED
BLOCK azure-pipelines.yml:69 A04 REMOVED_IMAGE_OR_TASK
BLOCK templates/notify-steps.yml:5 A02 UNTRUSTED_TEXT_IN_SCRIPT
REVIEW azure-pipelines.yml:17 A13 TEMPLATE_REPO_NOT_PINNED
REVIEW azure-pipelines.yml:24 A12 PRIVILEGED_CONTAINER
REVIEW azure-pipelines.yml:33 A20 PR_BUILD_ON_SELF_HOSTED_POOL
REVIEW azure-pipelines.yml:42 A11 PERSIST_CREDENTIALS
REVIEW azure-pipelines.yml:47 A15 DOWNLOAD_AND_RUN
REVIEW azure-pipelines.yml:49 A16 FREE_TEXT_INPUT_IN_SCRIPT
REVIEW azure-pipelines.yml:51 A18 RETIRING_TASK
REVIEW azure-pipelines.yml:56 A14 SECRET_ON_COMMAND_LINE
REVIEW azure-pipelines.yml:64 A17 DEPRECATED_HOSTED_IMAGE
REVIEW azure-pipelines.yml:76 A19 SPN_CREDENTIALS_IN_SCRIPT
(Each finding also has Why and Safer lines, as in the first one.)
Verdict: BLOCK. Checked 2 files, 2 with Azure Pipelines content.
This is a static check of file text. NOT FLAGGED is not a guarantee.
Connects securely to your tools. The creator never sees your data.
About this skill
A step that runs echo "Building $(Build.SourceVersionMessage)" looks harmless, but Azure Pipelines pastes the commit message into the script before bash starts, so a commit message with a quote or $( ) in it runs as a command on your agent, next to your secrets. A pipeline still on vmImage: windows-2019 stopped working when that image was removed on 31 December 2025, and CmdLine@1, PowerShell@1, Maven@2 and 70 other task versions are retired on 24 November 2026 and removed on 15 March 2027. This skill finds those and the other Azure Pipelines mistakes below, and tells you how to fix each one.
Your agent runs a scanner on azure-pipelines.yml, files in .azure-pipelines/ and .azuredevops/, and every YAML file whose top-level keys and steps are Azure Pipelines ones, so step, job, stage, variable and extends templates are checked too, including those kept under build/ that a pipeline includes. GitHub Actions, GitLab CI, Compose and Kubernetes files are skipped. It knows macro syntax $(var), template expressions ${{ }} and runtime expressions $[ ], treats values mapped through env: as safe, and follows variables built from untrusted values within a file. It never runs the pipeline or a script and never uses the network. It returns BLOCK, REVIEW, or NOT FLAGGED, with the file, line, reason, and a safer approach. Secrets are never printed.
What it checks
BLOCK
- A01: passwords, tokens and keys written as literals: variables or env entries with secret names, task inputs such as keystorePass, parameter defaults, passwords in URLs or connection strings, Bearer headers, --token flags, curl -u user:pass, docker/az login -p, sqlcmd -P, PowerShell $password = '...', -Password and ConvertTo-SecureString -AsPlainText, issecret logging commands, and known token formats (ghp_, npm_, glpat-, xox, AWS access key IDs, private key blocks)
- A02: $(Build.SourceVersionMessage) or $(System.PullRequest.SourceBranch), or a variable built from them, inside script text (script, bash, pwsh, powershell, inline task scripts and shell task arguments)
- A03: scripts fetched over plain HTTP or with certificate checks off (curl -k, wget --no-check-certificate, -SkipCertificateCheck) and piped into a shell or Invoke-Expression
- A04: removed images and tasks: retired Microsoft-hosted images (windows-2019, vs2017-win2016, ubuntu-16.04 to ubuntu-20.04, macOS-10.13 to macOS-13, and macOS-14 and ubuntu-22.04 from their removal dates), also through a matrix variable or parameter that vmImage uses, and from 15 March 2027 the retired task versions
REVIEW
- A11: checkout with persistCredentials: true
- A12: container jobs or service containers with --privileged, powerful capabilities, host namespaces, seccomp or AppArmor off, a docker.sock volume or mapDockerSocket: true
- A13: templates from GitHub, Bitbucket or another organization's repository resource with no ref or a branch ref (a tag or 40-character commit SHA counts as pinned)
- A14: $(System.AccessToken) or secret-named variables pasted into script text or task command-line arguments instead of env:
- A15: scripts fetched over HTTPS and run straight away (curl | sh, irm | iex, bash <(curl ...), iex "& { $(irm ...) }", [scriptblock]::Create((iwr ...)))
- A16: free-text runtime string parameters without a values list, or $(Build.RequestedFor) and the other requester names, inside script text
- A17: deprecated hosted images before their removal date: macOS-14 (23 November 2026) and ubuntu-22.04 (17 April 2027)
- A18: the 73 task versions on Microsoft's November 2026 retirement list, each shown with its replacement, until their removal on 15 March 2027
- A19: AzureCLI with addSpnToEnvironment: true
- A20: an explicit pr trigger in a pipeline that runs jobs on a self-hosted pool
What's included
- The scanner (Python 3.9+, standard library only)
- A rules guide with the reason, safer approach and Microsoft source for each rule
- A risky and a safe example (a build and release pipeline with a steps template), and the real report the scanner produces on the risky one
Limits It reads each YAML file as written. Values passed between files (template parameters, variables from other files, variable groups, UI and queue-time variables) are not followed, and settings made outside YAML are not seen: fork build and fork secret settings, required template checks, queue-time variable limits, shell argument validation, job authorization scope, branch policies and whether the repository is public. Branch names such as $(Build.SourceBranchName) are not flagged. In a whole-repository scan, a build/ folder is searched only for pipeline-named files and local templates, and folders such as dist, vendor and node_modules are skipped. Image and task lists follow Microsoft's pages as of 12 November 2026 for Azure DevOps Services. Classic and release pipelines defined in the web UI are not read. NOT FLAGGED is not a guarantee.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.
Static check of YAML text. Template parameters and variables from other files, variable groups, pipeline UI settings, fork build settings, required template checks and Azure DevOps Server schedules are not checked. NOT FLAGGED is not a guarantee.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 4 days ago
Permissions
Runs one local Python script in the terminal. It reads your project's files, writes nothing, and makes no network calls. The example project includes one clearly fake, inert token (not-a-real-sonar-token-for-examples) for the scanner to find. Nothing in it is run.