The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains no attempts to bypass safety filters or override system instructions. It includes robust 'Anti-Patterns' and 'Ethical Boundaries' sections that explicitly instruct the agent to avoid authoritative diagnosis or clinical therapy, which serves as a safety constraint rather than a malicious injection.
[DATA_EXFILTRATION]: No sensitive file paths, environment variables, or hardcoded credentials were detected. While the skill utilizes tools like WebFetch and search, it does so within the context of academic and psychological research, pointing to established institutions such as the IAAP and C.G. Jung Institute.
[REMOTE_CODE_EXECUTION]: There are no patterns suggesting the download or execution of remote scripts. The skill consists entirely of instructional markdown and reference material.
[OBFUSCATION]: No obfuscated code, base64-encoded instructions, zero-width characters, or homoglyph attacks were found. The extensive use of ASCII diagrams is transparent and used for pedagogical purposes.
[INDIRECT_PROMPT_INJECTION]: The skill uses mcp__firecrawl__firecrawl_search and WebFetch to gather external information. This presents a theoretical attack surface where malicious third-party content could influence the agent; however, the skill provides a highly structured internal context and clear clinical frameworks that reduce the likelihood of the agent obeying external instructions over its primary persona rules.

jungian-psychologist