The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The resource references/knowledge-base.md contains a URL to getmonetizely.com which has been flagged as a phishing domain by automated security scanners.
[DATA_EXFILTRATION]: Automated antivirus scans identified the file references/knowledge-base.md as infected with the MD:HttpRequest-inf signature, which is associated with malicious network activity or data exfiltration attempts.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface by design as it ingests and processes arbitrary user datasets and dashboard metadata.
Ingestion points: Dataset uploads and database connections mentioned in SKILL.md workflows.
Boundary markers: Absent. The instructions do not define boundaries or specific directives for the agent to ignore instructions found within processed data.
Capability inventory: Python script execution and SQL query execution are core capabilities for processing user data.
Sanitization: Absent. The skill does not mandate sanitization or validation of the data content before analysis.
[REMOTE_CODE_EXECUTION]: The skill's 'Hands-on Execution' workflow explicitly directs the agent to execute Python and SQL code based on external data, creating an execution surface for any malicious payloads embedded in that data.

data-analysis