The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute rg (ripgrep) to scan project files for vulnerability patterns and the CFR decompiler to process binary bytecode. These commands are executed locally within the agent's environment to fulfill auditing tasks.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project data including source code, configuration files, and compiled bytecode, presenting a surface for adversarial content to potentially influence the analysis.\n
Ingestion points: Project source code, binary .class files, and XML/Properties configuration files as defined in SKILL.md (Coverage Rules) and DESERIALIZATION_SINKS.md.\n
Boundary markers: There are no explicit markers or instructions provided to the agent to segregate instructions from data in the files it analyzes.\n
Capability inventory: The agent can execute system commands (rg) and run decompilation logic (CFR), which could be misused if influenced by malicious input files.\n
Sanitization: The instructions do not define any validation or sanitization procedures for the project files being audited.

java-deserialization-audit