The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration were detected. The skill instructions prioritize safe data handling and restrict the agent to a diagnostic, read-only role.- [PROMPT_INJECTION]: The skill processes untrusted external data (search term reports), creating a potential surface for indirect prompt injection. Ingestion points: The skill accepts pasted search-term data or live data via the Google Ads MCP. Boundary markers: The instructions do not include specific delimiters to isolate the processed data from instructions. Capability inventory: The skill is limited to read access and explicitly commands the agent to "Change nothing" and "Do not add negatives," preventing automated state changes. Sanitization: No input sanitization is mentioned, though the strict read-only scope minimizes potential impact.

search-term-miner