crowdin-context-cli

Pass

Audited by Gen Agent Trust Hub on Feb 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill documents the execution of the crowdin CLI tool to perform context management tasks. This involves running subprocesses like crowdin context download and crowdin context upload on the local system.
  • [CREDENTIALS_UNSAFE]: The documentation references authentication mechanisms, specifically the use of the -T, --token flag and configuration files (crowdin.yml) to provide Crowdin API tokens. While it describes how to use these credentials, it does not include hardcoded secrets.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface. It involves downloading strings and context from Crowdin (crowdin context download) for the AI agent to process and enrich. If the source content in Crowdin is attacker-controlled, it could contain instructions designed to influence the agent's behavior.
  • Ingestion points: Data enters the context via crowdin-context.jsonl files downloaded from the Crowdin API.
  • Boundary markers: The data is structured in JSONL format, but no specific prompt delimiters or instructions to ignore embedded commands are described.
  • Capability inventory: The skill allows file reading/writing and execution of the crowdin CLI.
  • Sanitization: No explicit sanitization or validation of the downloaded source text is mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 23, 2026, 02:38 PM