lynx-debug-info-remapping
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill workflow requires the agent to ingest and process external data files (
debug-info.jsonandmain-thread.js) provided by the user. This creates a surface where malicious instructions could be embedded within the source code or metadata to influence the agent's logic. - Ingestion points: The agent is instructed to ask for and read the content of the
debug-info.jsonandmain-thread.jsfiles (SKILL.md, Workflow Steps 2 and 5). - Boundary markers: No specific delimiters or "ignore instructions" warnings are utilized when the agent reads these files.
- Capability inventory: The skill utilizes shell command execution to run a Node.js script and has the capability to display file contents to the user.
- Sanitization: There is no evidence of sanitization for the content of the files before they are processed by the agent or the script.
- [DYNAMIC_EXECUTION]: The script
scripts/index.mjsuses the dynamicimport()function to load the debug information file from a path provided via command-line arguments. While the use of thetype: 'json'import attribute limits the scope to JSON data, loading modules from computed paths at runtime remains a dynamic execution risk factor. - [COMMAND_EXECUTION]: The skill relies on the agent executing a shell command (
node ...) using variables ($DEBUG_INFO, $function_id, $pc_index) derived from user input or project discovery. This presents a risk of command injection if the agent does not properly escape or quote these variables when invoking the shell.
Audit Metadata