The Agent Skills Directory

[SAFE]: The skill instructions and evaluation files do not contain any evidence of malicious patterns, such as obfuscation, credential exfiltration, or unauthorized remote code execution. References to Microsoft Agent Framework are consistent with legitimate development tools.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute a local Python script (test_harness.py) located within the skill's directory structure. These commands are used to simulate agent execution and validate plugin manifests, which aligns with the skill's stated purpose.
[PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by processing external skill directories and input files (--skill, --input) during simulation. However, this is an inherent functional requirement for a test harness, and the skill does not include any instructions that attempt to bypass safety filters or override agent behavior.

maf-adapter