date-calculator
Warn
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONMETADATA_POISONINGCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: Self-referential content detected in
date-calculator_audit_result_v2.json. This file contains a pre-generated, detailed security audit that explicitly assigns the skill a "PASS" verdict and high security scores. This is a sophisticated attempt to bias the security analysis process by providing a counterfeit authoritative safety assessment within the skill package itself. - [METADATA_POISONING]: Deceptive capability and dependency claims. The
SKILL.mdfile and the audit JSON document a--timezoneparameter and integration with thepytzlibrary. However, thescripts/main.pyimplementation lacks this parameter in its argument parser and does not import or use any timezone libraries. These claims mislead users and agents about the tool's actual functional boundaries and dependencies. - [COMMAND_EXECUTION]: Arbitrary file-write capability in
scripts/main.py. The script accepts a user-provided file path via the--outputparameter and writes its execution results to that location usingwith open(args.output, 'w', encoding='utf-8') as f: f.write(output). Without path sanitization, this allows the agent to overwrite files within the execution environment.
Audit Metadata