The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates on external files, which creates a standard attack surface for indirect prompt injection.
Ingestion points: File content is ingested using the Read and Grep tools as specified in SKILL.md.
Boundary markers: Absent; there are no specific instructions or delimiters provided to the agent to treat file content as untrusted data.
Capability inventory: The skill utilizes Bash for command execution and Write/Edit for filesystem modifications (SKILL.md).
Sanitization: Absent; the skill does not define validation or filtering logic for data read from the filesystem.
[SAFE]: The skill's primary operations involve legitimate built-in tools for file reading, writing, and searching.
[SAFE]: External tool references are limited to well-known developer utilities like pnpm and ripgrep.
[SAFE]: The persistence mechanism described (Memory Protocol) is a standard practice for agent state management and does not facilitate unauthorized access.

filesystem