skills/mukul975/anthropic-cybersecurity-skills/hunting-for-registry-run-key-persistence/Gen Agent Trust Hub
hunting-for-registry-run-key-persistence
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: Minor inconsistencies were identified in the skill metadata. The author is listed as 'mahipal' in SKILL.md, but the LICENSE file credits 'mukul975'. Additionally, the SKILL.md lists 'xml.etree.ElementTree' as a prerequisite, but the provided scripts/agent.py does not actually import or use this module for its analysis.
- [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for processing untrusted external data through its log analysis script. 1. Ingestion points: Sysmon logs and registry snapshots provided via the --input argument in scripts/agent.py. 2. Boundary markers: Absent. 3. Capability inventory: Local file system read and write access is utilized for processing inputs and saving findings. 4. Sanitization: The script uses standard Python libraries for data parsing and performs string comparisons and regular expression matching to identify threats without executing the content of the logs.
Audit Metadata