The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/verify_deploy.py uses subprocess.run to execute the system curl utility. This is used to verify that the deployed application's endpoints are responding correctly.
[EXTERNAL_DOWNLOADS]: The scripts/verify_deploy.py script performs network requests to external URLs (deployment endpoints) to verify status. This is a core part of its verification functionality but constitutes an external network interaction.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8):
Ingestion points: scripts/verify_deploy.py fetches and returns the first 500 characters of HTTP response bodies from remote URLs. scripts/inspect_webapp_deploy.py reads content from local repository files such as package.json and .env.example.
Boundary markers: No explicit delimiters or instructions to ignore embedded content are used when returning this data to the agent context in the script's JSON output.
Capability inventory: The skill provides the agent with the ability to modify infrastructure and repository state using az and gh CLI commands as outlined in scripts/render_plan.py and references/workflow.md.
Sanitization: Ingested HTTP content is truncated to 500 characters, which provides a minor reduction in risk but does not eliminate the potential for malicious instructions to influence the agent.

azure-webapp-deploy