The Agent Skills Directory

[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because its primary function is to audit untrusted Solidity code provided by users.
Ingestion points: Untrusted source code enters the agent context through user prompts or project files being analyzed, as described in Phase 3 of SKILL.md.
Boundary markers: The skill does not instruct the agent to use specific delimiters or to ignore natural language instructions that might be embedded in code comments or string literals.
Capability inventory: The skill authorizes the agent to execute shell commands for security tools like slither, forge, and aderyn (SKILL.md Phase 2).
Sanitization: There is no specific instruction for the agent to sanitize or filter the input code before processing it.
[COMMAND_EXECUTION]: The skill instructs the agent to perform automated analysis using standard security tools. In SKILL.md Phase 2, the agent is directed to run slither, forge, and aderyn if they are available in the environment. These tools are industry standard and appropriate for the skill's purpose.

solidity-security-audit