The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions utilize shell commands to set up temporary test environments using mktemp and mkdir. It also permits the execution of 'live commands' during the testing phase to verify the functionality of a target skill.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent to install small, low-risk utilities if they are necessary to satisfy the dependencies of the skill being tested. This is gated by safety rules requiring such installations to be reversible and disposable.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because its primary function is to ingest and analyze instructions from third-party 'target skills'.
Ingestion points: Target skill files including frontmatter, instructions, scripts, and examples are loaded into the agent's context (SKILL.md).
Boundary markers: No specific delimiters are defined to isolate external instructions from the meta-skill's own logic.
Capability inventory: The agent has access to file system operations, command execution, and potential network access through its toolset.
Sanitization: The instructions do not specify a requirement to sanitize or filter the content of the target skills before processing.

skill-test-drive