The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions guide the agent to construct and execute shell commands (specifically grep and find) using user-provided input ($ARGUMENTS). If the agent does not properly escape or sanitize these arguments before execution, a user could potentially supply shell metacharacters (e.g., semicolons, backticks, or pipes) to execute arbitrary commands on the system.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by ingesting and processing data from the local Maven repository (~/.m2/repository) and the current project directory.
Ingestion points: The agent reads class names and JAR paths from ~/.cache/quarkusdev-skills/class-index.txt (populated by jar tf) and searches the local filesystem using find in SKILL.md (Step 4).
Boundary markers: The skill does not provide specific delimiters or instructions to the agent to treat search results as untrusted data or to ignore embedded instructions within those results.
Capability inventory: The agent has the capability to execute shell commands (bash), read/write files, and execute the included build-class-index.sh script.
Sanitization: No sanitization or validation logic is defined to check for malicious filenames or crafted class paths that might contain instructions designed to manipulate the agent's behavior during result presentation.

java-classpath-search