The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests and analyzes data from untrusted codebases. . Ingestion points: The skill uses sub-agents to scan all files within the project root, including package configurations, schema definitions, and documentation files (SKILL.md). . Boundary markers: The skill relies on natural language instructions for sub-agents (e.g., "Do NOT modify any files") which can be targeted for bypass by malicious instructions within scanned files (SKILL.md). . Capability inventory: The skill utilizes the Agent tool and shell commands (bash) to perform file system operations and generate summaries (SKILL.md). . Sanitization: There is no evidence of content validation or escaping of ingested data before it is processed by the agent system (SKILL.md).

import