The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes instructions to ingest untrusted data from external sources, creating a potential injection surface.
Ingestion points: The agent is instructed to read output from azd env get-value WEB_ENDPOINT and az containerapp logs show in SKILL.md to verify deployment status.
Boundary markers: There are no delimiters or instructions to ignore embedded commands within the log data.
Capability inventory: The skill possesses extensive capabilities including shell command execution (azd, az, dotnet, npm) and local script execution (node) as seen in SKILL.md.
Sanitization: No sanitization or validation of the log content or environment variables is performed before processing.
[DYNAMIC_EXECUTION]: The instructions require the agent to execute a local JavaScript file (deployment/scripts/smoke-test.js) using Node.js to validate the environment. This represents the execution of code from a static path within the project structure. Evidence found in SKILL.md.
[COMMAND_EXECUTION]: The skill performs multiple high-privilege administrative actions, including infrastructure provisioning (azd up), resource teardown (azd down --force --purge), and identity management (az ad app federated-credential create). While consistent with its stated purpose of deployment validation, these actions involve significant interaction with the host system and cloud environment. Evidence found in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill fetches external binaries during the prerequisite phase using npx playwright install chromium. This downloads the Chromium browser from official sources for testing purposes.

validating-deployments