ai-automation-workflows
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The 'Data Processing Pipeline' section in
SKILL.mddemonstrates an indirect prompt injection surface by interpolating raw file content into an LLM prompt. - Ingestion points: The script reads all
.txtfiles from the./data/rawdirectory. - Boundary markers: Absent; the content of the files is directly appended to the prompt instruction without delimiters or guardrails.
- Capability inventory: The skill uses
belt app runto execute AI operations and shell redirection (>) to write outputs to the file system. - Sanitization: Absent; the raw output of
cat $fileis placed into the JSON input for the model. - [COMMAND_EXECUTION]: The skill provides extensive examples for executing shell commands, bash scripts, and Python scripts to orchestrate AI workflows using the
beltCLI. - [EXTERNAL_DOWNLOADS]: The documentation references external installation scripts and additional skills hosted on the vendor's GitHub repository (
github.com/inference-sh/skills).
Audit Metadata