The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting and analyzing untrusted external content, specifically the target codebase files and architectural documentation such as sast/architecture.md. These serve as ingestion points where adversarial instructions could be embedded. The instructions do not define explicit boundary markers or sanitization logic to mitigate potential prompt injection from the analyzed source code. The skill's capability inventory includes file system read/write operations within the analysis directory and the orchestration of subagents.
[COMMAND_EXECUTION]: The verification instructions guide the agent to generate testing documentation that includes sqlmap and curl commands. This pattern involves the construction of shell-like command strings based on the vulnerabilities identified during the analysis phase.

sast-sqli