The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to locate and execute shell commands described in the project's documentation. In 'SKILL.md', the agent is instructed to 'run its documented checks' and 'run the exact setup, reference, and check commands' specified in a pattern's README.md file.
[DYNAMIC_EXECUTION]: The instructions require the agent to dynamically prepare and run codebases found at runtime. This includes installing prerequisites and executing reference implementations as part of the 'learner journey' verification process described in 'SKILL.md'.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and acts upon instructions stored in the project repository. Ingestion points: Pattern 'README.md' files, manifests, and implementation code located in the 'patterns/' subdirectories. Boundary markers: The instructions tell the agent to 'read each relevant README... in full' and compare patterns against requirements, but there are no explicit instructions to disregard or escape malicious embedded shell commands. Capability inventory: The agent utilizes shell access to install packages, run reference implementations, and execute automated checks. Sanitization: There is no evidence of command sanitization or validation; the agent is simply instructed to run the 'exact' commands found in the documentation.

pattern