The Agent Skills Directory

[COMMAND_EXECUTION]: The test script scripts/test_score_content.py invokes the main scoring script via subprocess.run. It uses the current Python interpreter (sys.executable) to run the local scripts/score_content.py file with specific test arguments. This is a controlled execution for verification purposes and does not introduce a security vulnerability.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted text (marketing drafts) through regex matching and arithmetic scoring. While this presents an ingestion surface for potentially malicious text, the skill does not perform any sensitive operations based on the input. It lacks network capabilities and only writes to a dedicated local memory file.
[SAFE]: The skill reads from and writes to a local JSON file (.content-memory.json by default) to maintain a persistent store of rejected content patterns. This behavior is documented and constrained to the paths provided by the user or the agent. No evidence of unauthorized data exfiltration or access to sensitive system files was found.

content-scorer