The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local bash scripts located in the plugin's own scripts directory to scan transcripts and process behavioral data. It also uses python3 for minor tasks like JSON parsing and path resolution, which are restricted to handling the skill's own internal data files.\n- [EXTERNAL_DOWNLOADS]: No external network activity or script downloads were detected. All processing logic and resources are self-contained within the skill distribution.\n- [PROMPT_INJECTION]: The skill ingests untrusted transcript data as part of its core functionality. It mitigates indirect prompt injection risks by using highly restrictive synthesis templates that enforce a strict JSON output format and prohibit interpretive, evaluative, or subjective language. This design ensures that the agent's summary is based on factual patterns rather than potentially malicious instructions embedded in the analyzed data.\n- [DATA_EXFILTRATION]: Although the skill accesses sensitive local transcript data, it contains no mechanisms for external network communication or data exfiltration. All generated reports and persona data are stored locally within the user's project environment.

whoami