The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest external data to drive its primary functionality, creating a potential vector for indirect instructions.
Ingestion points: The skill reads test-cases.md, infra-decision-table.md, and fixture files in the testdata/ directory (Step 1, Step 2, and Step 3).
Boundary markers: The skill implements an integrity check in Step 1, requiring that input files are byte-identical to a known baseline commit.
Capability inventory: The skill generates code based on these inputs and executes it using system build tools such as mvn, gradle, dotnet, pytest, and go (Step 6).
Sanitization: The workflow includes a 'Pre-generation trigger audit' (Step 5) and a 'Pre-run validation checklist' (Step 6) to ensure the generated code conforms to strict safety and architectural rules.
Command Execution and Dynamic Code Generation: The skill generates *PostMigrationIT source files and subsequently executes them using project-specific build runners. This pattern involves the runtime execution of dynamically generated code. The skill mitigates risks by enforcing strict mapping rules (e.g., 1:1 mapping of test cases) and forbidding the use of internal application services or private helpers as test triggers.
Sensitive Configuration and Credential Access: The skill interacts with environment-specific configuration files (.env, application-integrationtest.yml, appsettings.IntegrationTest.json) and Azure CLI tools (az login, DefaultAzureCredential). While these are necessary for integration testing against cloud resources, they involve the handling of sensitive environment identifiers and authentication principals. The skill follows best practices by forbidding hardcoded secrets and requiring that credentials be resolved via standard identity providers or environment variables.

verify-test-baseline