The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to download and execute an installation script for the Azure CLI from a well-known Microsoft domain.
Evidence: curl -sL https://aka.ms/InstallAzureCLIDeb | sudo bash in SKILL.md.
[PRIVILEGE_ESCALATION]: The skill suggests using elevated privileges to install system-level tooling.
Evidence: Use of sudo bash for the Linux installation command in SKILL.md.
[COMMAND_EXECUTION]: The skill contains numerous patterns for executing shell commands, including network operations for sending notifications and downloading backups.
Evidence: Use of curl -X POST "$slack_webhook" and curl -L "$BACKUP_URL" in references/workflows-and-patterns.md.
[INDIRECT_PROMPT_INJECTION]: The skill provides capabilities to ingest and process untrusted data from external Azure DevOps resources which could contain malicious instructions.
Ingestion points: Commands such as az boards work-item show, az repos pr show, and az pipelines runs list (documented in references/boards-and-iterations.md, references/repos-and-prs.md, and references/pipelines-and-builds.md).
Boundary markers: No explicit boundary markers or instruction-ignore warnings are present when reading external content.
Capability inventory: The skill has extensive write capabilities including az repos pr create, az pipelines run, and az boards work-item update, as well as potential network exfiltration via curl notifications.
Sanitization: There is no evidence of sanitization for external content before it is processed by the agent.

azure-devops-cli