The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies user-provided code files as primary context for its operations, creating an indirect prompt injection surface.\n- Ingestion points: Workflow step 1 instructs the agent to read relevant files before editing.\n- Boundary markers: None specified in the instructions to separate external file content from the agent's core instructions.\n- Capability inventory: Includes file reading, file modification, and command execution for verification (e.g., linting, smoke tests) as part of the implementation process.\n- Sanitization: No specific sanitization or validation of the ingested code content is defined within the skill instructions.\n- Assessment: This surface is characteristic of coding-oriented skills and represents standard behavior for the task described.

ponytail