debugging-code
Fail
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires downloading and installing a CLI tool from a community GitHub repository (
AlmogBaku/debug-skill). Thescripts/install-dap.shscript automates the download of these binaries usingcurl. - [COMMAND_EXECUTION]: Instructions recommend running a local shell script (
bash scripts/install-dap.sh) to install system-level tools. This script also suggests modifying user shell configuration files (such as.bashrcor.zshrc) to update the system PATH. - [REMOTE_CODE_EXECUTION]: The primary function of the skill is to launch and debug active processes, which involves executing code. The
daptool uses various language adapters to perform remote and local code execution for debugging purposes. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests live program output (stdout and local variables) which could be manipulated by the code being debugged.
- Ingestion points:
dapcommand output containing program state and stdout. - Boundary markers: Not utilized; the agent is instructed to process output directly.
- Capability inventory: Shell command execution via the
dapCLI and arbitrary expression evaluation viadap eval. - Sanitization: Not implemented.
- [EXTERNAL_DOWNLOADS]: Fetches the
js-debugadapter from Microsoft's official GitHub repository.
Recommendations
- AI detected serious security threats
Audit Metadata