The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables the agent to read and analyze arbitrary source code and documentation files. This creates a surface for indirect prompt injection, where malicious instructions hidden in the files (such as in code comments or markdown documentation) could influence the agent's behavior.
Ingestion points: Data enters the agent's context through the read, search, meta_search, and diff tools when processing files like SKILL.md or repository source code.
Boundary markers: The skill does not provide instructions for using boundary markers or explicit prompts to ignore instructions found within the processed data.
Capability inventory: The agent has access to powerful write and batch tools (edit, patch, create, replace, batch) that could be targeted by an injection attack.
Sanitization: The skill lacks instructions for sanitizing or escaping the content retrieved from external files before it is used to inform further agent actions.

codedbpro