The Agent Skills Directory

[SAFE]: No malicious patterns, such as direct prompt injection, credential harvesting, or unauthorized remote code execution, were identified in the skill instructions or metadata.
[COMMAND_EXECUTION]: The skill utilizes the ToolSearch mechanism to discover and load deferred MCP tools (ouroboros_session_status and ouroboros_measure_drift). These are specialized analytical tools within the Ouroboros framework used for session management and are invoked with parameters like session_id and current_output.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external data (current_output) through its measurement tools.
Ingestion points: Untrusted data enters the context via the current_output argument in the ouroboros_measure_drift tool call, which represents current execution state or file contents.
Boundary markers: The instructions do not define specific boundary markers (e.g., XML tags or delimiters) to isolate the untrusted content from the tool's core logic.
Capability inventory: The skill is limited to checking status and measuring drift; it does not have the capability to write files, execute arbitrary shell commands, or perform network requests outside the MCP framework.
Sanitization: There is no evidence of content sanitization or instruction filtering before the data is passed to the drift measurement tool. However, this is a standard operational surface for analytical tools and does not indicate malicious intent.

status