The Agent Skills Directory

[COMMAND_EXECUTION]: The engine script scripts/portfolio.py executes system commands via subprocess.run. It specifically calls the platform-native binary openclaw to retrieve runtime status and strategy registry information. The execution is limited to predefined subcommands (senpi runtime list, senpi status) and uses specific identifiers retrieved from the local registry, avoiding shell injection risks.
[EXTERNAL_DOWNLOADS]: The skill fetches a strategy enrichment catalog (catalog.json) from the vendor's official GitHub repository (github.com/Senpi-ai). This download is used purely for data enrichment (archetypes and theses) and originates from the trusted author's own infrastructure.
[PROMPT_INJECTION]: The instructions in SKILL.md guide the agent to provide a professional narrative by presenting findings as established facts rather than narrating technical lookups. While flagged by automated detectors for concealment, these instructions are benign UX/persona guidance designed to improve conversational quality rather than hide malicious actions.
[SAFE]: The skill demonstrates secure practices, including the use of environment variables for authentication tokens and implementing multi-step data reconciliation to ensure accuracy. The analysis of the Python scripts and markdown instructions revealed no malicious patterns or data exfiltration attempts.

senpi-portfolio