The Agent Skills Directory

[SAFE]: The skill instructions are focused on providing coaching feedback and improving developer habits. It contains strong safety directives to protect sensitive data, explicitly stating 'Do not expose secrets, tokens, private keys, or sensitive personal data' in both the main skill and subagent prompts.\n- [COMMAND_EXECUTION]: The skill mentions the use of agentlytics-setup and agentlytics --join localhost:4638, which are documented setup requirements for the vendor's service. These commands target a local instance and are consistent with the skill's stated purpose of analyzing Agentlytics history.\n- [DATA_EXPOSURE]: The skill processes session transcripts which may contain sensitive logs. However, the author has implemented explicit redaction rules and warns against quoting sensitive fragments, using paraphrasing instead to maintain audit integrity without compromising security.\n- [INDIRECT_PROMPT_INJECTION]: As the skill ingests external data (session transcripts), it is theoretically susceptible to indirect injection. This is mitigated by the instruction to use a structured rubric with a subagent, the lack of dangerous file-write or system-modifying capabilities, and the mandatory requirement for human-in-the-loop setup validation.

agentlytics-session-analysis