bash-shell

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill provides examples of sourcing external configuration files using a variable path (e.g., source "${CONFIG_FILE}" in examples/advanced-script.sh and references/common-patterns.md). This pattern allows for the execution of arbitrary shell code contained within the sourced file.
  • [COMMAND_EXECUTION]: In examples/advanced-script.sh, an unquoted here-document (<<EOF) is used to incorporate file content (line 208). Because the here-doc delimiter is not quoted, the shell will perform command substitution and variable expansion on the content of the input file ($(cat "${input_file}")) during execution. This could lead to arbitrary command execution if the script processes a file containing malicious shell sequences.
  • [EXTERNAL_DOWNLOADS]: The examples/library-example.sh file includes a utility function lib::download_file that uses curl to download content from a remote URL. While the function includes URL validation and optional checksum verification, it facilitates fetching external data into the agent's environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines patterns for scripts that ingest and process external data (e.g., reading file contents in advanced-script.sh and basic-script.sh). This establishes an attack surface where malicious instructions embedded in processed files could attempt to influence the agent's behavior.
  • Ingestion points: examples/advanced-script.sh (processes files in INPUT_FILES array), examples/basic-script.sh (reads input_file).
  • Boundary markers: The provided script templates do not implement specific delimiters or instructions for the agent to ignore potentially malicious embedded content within the processed data.
  • Capability inventory: The included scripts demonstrate capabilities for file system operations (rm, cp, mkdir), network access (curl), and dynamic execution (source).
  • Sanitization: The skill provides utility functions in examples/library-example.sh for URL validation and checksum verification to mitigate download risks, but does not provide specific sanitization for data processed within the shell scripts themselves.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 08:19 PM