skills/mukul975/anthropic-cybersecurity-skills/detecting-malicious-scheduled-tasks-with-sysmon/Gen Agent Trust Hub
detecting-malicious-scheduled-tasks-with-sysmon
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The analysis script
scripts/agent.pyprocesses external XML files provided by the user. A crafted input file could theoretically be used to influence the output of the analysis or cause processing errors. \n - Ingestion points: The
parse_evtx_xmlfunction inscripts/agent.pyaccepts a file path for XML event data via the--evtx-xmlargument.\n - Boundary markers: None present; the script assumes the integrity of the provided log file.\n
- Capability inventory: The script is limited to local file reading and writing results; it has no network or shell execution capabilities.\n
- Sanitization: No XML entity sanitization is performed on the input file.\n- [DYNAMIC_EXECUTION]: The script uses the standard
xml.etree.ElementTree.parsemethod to process event logs. While necessary for the skill's function, it is a form of dynamic data processing that lacks specific protections against XML External Entity (XXE) vulnerabilities.\n- [METADATA_POISONING]: There is a minor inconsistency between the author name 'mahipal' in the YAML frontmatter and the copyright holder 'mukul975' in the license file. This appears to be a documentation artifact and is not considered deceptive.
Audit Metadata