The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads external codebase content and translates it into executable test specifications without explicit sanitization or boundary markers.
Ingestion points: The agent investigates the codebase using glob and grep to derive context for integration tests as defined in the 'Core Process' (SKILL.md).
Boundary markers: The skill does not define delimiters or provide warnings to the agent to ignore instructions embedded within the codebase files it processes.
Capability inventory: The skill utilizes subagents with the capability to execute arbitrary shell commands, perform HTTP requests, and access the filesystem to verify test expectations.
Sanitization: There is no evidence of a sanitization or validation layer to prevent malicious code patterns in the source files from being included in the generated test 'Prerequisites' or 'Steps'.
[COMMAND_EXECUTION]: The skill's primary function involves defining and executing shell commands through subagents to perform system-level validation.
Evidence: The 'Core Process' and 'Quick Reference' sections in SKILL.md explicitly instruct the agent to document and execute shell commands for environment setup and test verification.
Evidence: The provided example auth-integration.md demonstrates the use of commands like docker compose up, npm run migrate, and sqlite3 to interact with the host environment.

agent-integration-testing