external-cannbot-ops-ascendc-crash-debug

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for using standard system commands such as ulimit to enable core dumps and gdb for analyzing process crashes. These operations are restricted to debugging local executables and are consistent with the skill's primary function.
  • [EXTERNAL_DOWNLOADS]: The skill references official Huawei Ascend documentation and a dedicated repository for CANN development skills. These references are documented neutrally as they point to established technical resources relevant to the developer workflow.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes a log parsing script, scripts/parse_plog.py, which processes external log data generated by the Ascend hardware stack. While this involves ingesting external data, the risk is minimal given the source and processing method.
  • Ingestion points: The script reads plog files from the ~/ascend/log/debug/plog directory, which are generated by the Ascend C runtime.
  • Boundary markers: The script does not use specific delimiters but performs pattern matching to categorize system errors.
  • Capability inventory: The script is limited to local file reading and printing summaries to the console; it contains no network or file-writing capabilities.
  • Sanitization: The script uses fixed regular expressions to filter for known error signals and truncates output snippets to 200 characters, mitigating the display of large or potentially manipulative log payloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 03:53 PM
Security Audit — agent-trust-hub — external-cannbot-ops-ascendc-crash-debug