The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted source code as input, which represents a surface for indirect prompt injection. Malicious instructions could be embedded in code comments or strings to influence the agent's behavior during analysis.
Ingestion points: The skill reads user project files using the Read, Grep, and Glob tools.
Boundary markers: The instructions do not define delimiters or specific 'ignore' instructions to isolate the user-provided code from the agent's internal logic.
Capability inventory: Includes the Bash, Read, Grep, and Glob tools, which are used to inspect and analyze the local environment and code.
Sanitization: No explicit sanitization or validation of the input code is described in the detection instructions.

codeprobe-solid