The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use a bash command string (find <corpus-dir> -maxdepth 1 ... | wc -l) to count files. This involves interpolating a user-provided directory path directly into a shell command, which could lead to command injection if the agent does not properly sanitize the input.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of a user's writing corpus (PDF and TeX files).
Ingestion points: The find command identifies files in a user-specified directory, and subagents are spawned to read these files individually.
Boundary markers: The instructions do not specify any boundary markers or instructions to the subagents to ignore potentially malicious embedded content within the corpus documents.
Capability inventory: The skill utilizes Read, Grep, Glob, Write, Bash, and Agent (subagents). A successful injection in a source document could potentially influence the subagent's output or the final synthesis written to voice-profile.md.
Sanitization: There are no explicit sanitization or filtering steps defined for the content extracted from the corpus before it is processed by the subagents or synthesized by the main agent.

voice-profile