skills/mukul975/anthropic-cybersecurity-skills/building-soc-metrics-and-kpi-tracking/Gen Agent Trust Hub
building-soc-metrics-and-kpi-tracking
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external SIEM and incident ticketing systems, creating a surface for indirect prompt injection findings.
- Ingestion points: Ingests alert disposition, incident timestamps, and lifecycle data from Splunk and ticketing systems like ServiceNow or Jira as specified in SKILL.md and scripts/agent.py.
- Boundary markers: No delimiters or instructional warnings are implemented to separate ingested data from the report output.
- Capability inventory: The scripts/agent.py script has the capability to write aggregated data to local JSON files.
- Sanitization: The script does not perform sanitization or validation of the external data before incorporating it into the final performance report.
- [DATA_EXFILTRATION]: The scripts/agent.py script includes functionality to bypass TLS certificate verification for network requests to the Splunk API.
- Evidence: The script checks the SKIP_TLS_VERIFY environment variable to disable SSL verification in the requests library calls, which could facilitate data interception via man-in-the-middle attacks.
- [CREDENTIALS_UNSAFE]: The Python agent script is designed to accept sensitive credentials as plaintext command-line arguments.
- Evidence: The main function in scripts/agent.py uses argparse to define a --password parameter, which can result in the exposure of Splunk credentials in process monitoring tools, process logs, or shell history.
Audit Metadata