The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses sensitive personal data within the $BRAIN_ROOT/topics/ directory to provide conversational context. This access is moderated by a topicId parameter which is constrained by a strict regex (^[a-z0-9][a-z0-9-]*$) to mitigate the risk of directory traversal and unauthorized file exposure.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its ability to process untrusted external data.
Ingestion points: External data is ingested through the read_article tool (which processes user-provided links), web_search results, and local topic files in SKILL.md.
Boundary markers: No explicit delimitation or boundary instructions are provided within the skill's markdown to distinguish untrusted content from system instructions.
Capability inventory: The skill utilizes search_brain, read_brain_page, read_article, and web_search. These tools are restricted to read-only operations in the standard configuration.
Sanitization: While internal identifiers are validated via regex, the skill lacks specific sanitization or filtering logic for content retrieved from external web sources.

voice-persona-mars