The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requests access to the Bash tool. The instructions in the SETUP phase specifically direct the agent to use this capability to configure the environment, such as setting the solc version based on the content of a foundry.toml file.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted Solidity files and directories provided as arguments. Maliciously crafted content within these files (e.g., within comments or string literals) could attempt to override the auditor's protocols or influence its findings.
Ingestion points: Smart contract files read via Read, Glob, and Grep tools.
Boundary markers: None explicitly defined to separate contract code from instructions, making it possible for the LLM to misinterpret code comments as authoritative instructions.
Capability inventory: The skill has access to Bash, file system operations, and several specialized auditing tools (run-slither, run-aderyn, etc.).
Sanitization: No explicit sanitization or instructions to disregard embedded commands in audited files are present.

security-auditor