skills/maxgfr/crible/crible-cli/Gen Agent Trust Hub

crible-cli

Pass

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructions include installing and running a CLI tool directly from a remote GitHub repository (github.com/maxgfr/crible) using uv tool install and uvx. Since this repository is owned by the skill's author ('maxgfr'), this is the standard and intended method for deploying the vendor's software.
  • [EXTERNAL_DOWNLOADS]: The skill performs several external data operations, such as pulling published datasets from GitHub (bootstrap), downloading financial data from Hugging Face (import-prices huggingface), and fetching worldwide stock data from Stooq (stooq-download). These actions are aligned with the skill's primary purpose of providing stock market fundamentals and pricing data.
  • [COMMAND_EXECUTION]: The skill is centered around executing shell commands via the uv package manager and the crible binary. These commands manage local data directories, run screening queries, and initiate background crawling processes.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted financial data from multiple external sources, including EDGAR, ESEF, and third-party price providers.
  • Ingestion points: Data is fetched via bootstrap, ingest, and refresh commands from external APIs and websites.
  • Boundary markers: The crible mcp interface provides a read-only surface and uses a strict Domain Specific Language (DSL) for queries, which limits the interaction to structured data fields.
  • Capability inventory: The CLI tool has file-writing capabilities (for database storage and exports) and network access (for crawling), but these are isolated from the LLM's direct prompt interpolation.
  • Sanitization: The tool uses a whitelist-based DSL (crible fields) for its screening queries, ensuring that only recognized financial metrics are processed, which mitigates the risk of malicious content in the data influencing the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 8, 2026, 03:06 PM