The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs audit logging to AWS S3 and DynamoDB to maintain a traceable record of remediation actions. These operations use official AWS service endpoints and are consistent with standard security auditing practices for incident response tools.
[DATA_EXPOSURE]: Manages a local configuration file in the user's home directory (~/.mcp-quarantine.jsonl) to store the tool blocklist. This access is localized to the tool's primary function and does not involve the collection or exposure of sensitive user credentials or system secrets.
[PROMPT_INJECTION]: The skill ingests external data from OCSF 1.8 Detection Findings. It mitigates potential injection risks by applying a deny-list for protected infrastructure tool prefixes (e.g., mcp_, system_) and sanitizing tool names before they are used in filesystem or storage paths.
Ingestion points: Findings are loaded from standard input or files into JSON objects in src/handler.py.
Boundary markers: None used during data processing.
Capability inventory: Local filesystem writes (appending to a blocklist file) and AWS cloud storage operations.
Sanitization: Implements is_protected_tool prefix checking and _safe_path_component filtering for tool identifiers.

remediate-mcp-tool-quarantine