The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute shell commands to compile and verify the project. This involves running ./mvnw compile, mvn compile, ./mvnw clean verify, and mvn clean verify. Executing these commands, especially the Maven wrapper (./mvnw), runs project-specific scripts that could potentially perform unintended actions if the project environment is untrusted.
[DYNAMIC_EXECUTION]: The skill follows a workflow where it modifies Java source code based on design principles and immediately executes the resulting code via the Maven lifecycle. This automated 'modify-then-execute' loop is a form of dynamic execution of generated content.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted data (user Java code) and has high-impact capabilities (file modification and shell execution).
Ingestion points: The agent reads and inspects user-provided Java source files as specified in the workflow of SKILL.md and references/122-java-type-design.md.
Boundary markers: Absent. There are no instructions provided to the agent to use delimiters or specific safety guards when reading the contents of the Java files to distinguish between code and potential malicious instructions.
Capability inventory: The skill has the capability to write to the file system (refactoring code) and execute shell commands (mvn, ./mvnw) as documented in SKILL.md.
Sanitization: Absent. There is no evidence of sanitization, filtering, or validation of the ingested code content before the agent processes it or incorporates it into its reasoning.

122-java-type-design