The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and analyze untrusted external data.
Ingestion points: Processes user-provided CVs and professional profiles (e.g., context/mentorship/profile/professional-profile.md) for career assessment and skill mapping.
Boundary markers: Absent. The instructions do not specify the use of delimiters or warnings to ignore instructions embedded within the processed CVs.
Capability inventory: The agent has access to Bash, WebSearch, Write, and Edit tools as defined in the frontmatter.
Sanitization: Absent. There is no mention of filtering, escaping, or validating the content of analyzed documents.
[DATA_EXFILTRATION]: The skill handles sensitive PII, including career goals, session notes, and professional resumes stored in context/mentorship/. The availability of WebSearch and Bash tools provides a technical path for this data to be sent to external domains if the agent's logic is subverted via injection.
[COMMAND_EXECUTION]: The frontmatter explicitly allows the Bash tool. While no malicious commands are hardcoded, this increases the potential impact of an successful prompt injection attack.

product-mentor