doppler-bench

Pass

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill defines several benchmark workflows that execute local scripts using node and npm. These include performance comparisons between different engines (tools/compare-engines.js) and vendor-specific benchmark runners (benchmarks/runners/transformersjs-bench.js).
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8) by interpolating untrusted data, such as model identifiers and configuration parameters, into shell commands.
  • Ingestion points: Use of MODEL_ID placeholders and complex JSON configuration strings in SKILL.md that are passed as command-line arguments to the benchmark tools.
  • Boundary markers: No explicit boundary markers or instruction-guarding delimiters are identified in the command templates.
  • Capability inventory: The skill facilitates the execution of local JavaScript tools (tools/vendor-bench.js, tools/compare-engines.js) and npm-defined tasks (npm run bench) which interact with the system's compute resources.
  • Sanitization: No explicit sanitization or validation of the provided configuration JSON is visible within the skill definition, relying on the underlying tools to handle potentially malformed or malicious inputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 12, 2026, 11:44 PM