The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a testing script (tests/test_grade_issue_spot.py) and a grader script (scripts/grade_issue_spot.py) that use the subprocess module. Technical evaluation confirms this is used exclusively for automated structural validation of the skill's own generated output during development and testing. The execution is restricted to calling the local Python interpreter on the internal grader script with static file paths as arguments, posing no risk to the host system.
[DATA_EXFILTRATION]: The skill explicitly operates as a pure-LLM workflow with no external fetches or network requirements. Analysis of the protocols and scripts confirms there are no network-capable commands (e.g., curl, wget) or patterns that could be used for data exfiltration.
[PROMPT_INJECTION]: The instructions and protocols maintain strict task focus. They implement mandatory safety checks, such as an escalation override for high-risk scenarios and mandatory legal disclaimers. No override or bypass instructions were detected.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied fact patterns. While this is an ingestion surface, the processing pipeline is entirely instruction-based and does not involve dynamic code generation or execution based on the user data. The analysis identifies this as a standard functional requirement rather than a vulnerability.

legal-issue-spot