The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions mandate that the agent re-read the target markdown file from disk before every write operation to preserve user-made changes. This ingestion of external data creates an attack surface where malicious instructions embedded in the file could influence the agent's future behavior or interview style. The skill lacks explicit boundary markers or sanitization logic to differentiate between stored fragments and potential instructions.
[COMMAND_EXECUTION]: The skill requires the ability to interact with the local filesystem to read, append, and edit markdown documents. It prompts the user for a storage path and performs silent background writes, which represents a capability that could be misused if the agent is directed to sensitive file paths.

writing-fragments