The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard local utilities to inspect project dependencies.
It executes unzip -l and grep to locate class files within JAR archives.
It uses javap -c -p to disassemble bytecode for analysis. These commands are executed locally on files provided within the target project's root.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code and library bytecode to generate configurations, creating a vulnerability surface where malicious code comments or identifiers could influence the agent's behavior.
Ingestion points: The agent reads source code directly from the project-root and library JARs located in .opentaint/project/dependencies/.
Boundary markers: No explicit markers or instructions are provided to the agent to treat source code content as data rather than instructions.
Capability inventory: The agent has the ability to execute shell commands (unzip, javap) and write configuration files (.yaml) to the project directory.
Sanitization: There is no evidence of sanitization or filtering of the source code content before it is processed by the agent.

create-pass-through-approximation