The Agent Skills Directory

[SAFE]: No security issues were detected. The skill instructions and scripts perform benign data retrieval and formatting.
[COMMAND_EXECUTION]: The skill uses a local script scripts/joke.py to fetch joke content. The script uses standard libraries like argparse, json, and random to process a local file references/jokes.json. There is no evidence of command injection or unsafe subprocess execution.
[DATA_EXFILTRATION]: No network operations or access to sensitive local file paths (like credentials or SSH keys) were found. The skill only interacts with its own bundled data and scripts.
[PROMPT_INJECTION]: The instructions in SKILL.md are focused on contextual delivery and "Iron Laws" (e.g., one joke per session, don't explain the punchline) which actually serve as guardrails for agent behavior rather than attempting to bypass safety filters.
[REMOTE_CODE_EXECUTION]: No patterns of downloading and executing remote code were identified. All dependencies are standard Python built-in modules.

nerdy-jokes