The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data including source code and security finding descriptions to perform its triage logic. This creates a surface for indirect prompt injection where adversarial content within the analyzed files (e.g., malicious code comments or specially crafted finding text) could attempt to influence the agent's final verdict.\n
Ingestion points: The agent reads from project directories (src/**, app/**, etc.) and processes original finding/question inputs.\n
Boundary markers: The skill instructions do not specify the use of clear delimiters or instructions to ignore embedded commands within the analyzed data.\n
Capability inventory: The skill utilizes Read, Grep, Bash, and Glob tools to perform verification, which could be misused if the agent's logic is subverted.\n
Sanitization: No explicit sanitization or validation of the external content is defined before it is interpolated into the triage prompts.

skeptical-triage