devup-ui
Fail
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The helper script
scripts/install.shcontains a shell command injection vulnerability in theinstall_skillfunction. The script iterates over theAGENTSenvironment variable and appends its values unquoted to a variableAGENT_FLAGS, which is then executed unquoted inside annpxcommand:npx skills add "$REPO_URL" --skill "$SKILL" $GLOBAL_FLAG $AGENT_FLAGS. This allows for arbitrary command execution if shell metacharacters (e.g., semicolons, backticks, or command substitutions) are present in theAGENTSinput.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of several Node.js packages under the@devup-uiscope (e.g.,@devup-ui/react,@devup-ui/next-plugin) through standard package managers. These are external dependencies required for the library's functionality.\n- [REMOTE_CODE_EXECUTION]: The instructions and scripts promote the use ofnpx skills addtargeting a remote GitHub repository (https://github.com/akillness/jeo-skills) to install the skill as a plugin. This mechanism involves fetching and potentially executing code from an external source.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines an intake process that gathers user-supplied configuration parameters, which are then passed to execution environments (shell scripts). This creates an attack surface where malicious input could be used to exploit the identified command injection vulnerability.\n - Ingestion points: The intake packet described in
SKILL.md(Step 1) and environment variables (AGENTS,BUNDLER,PROJECT,PKG_MANAGER) utilized byscripts/install.sh.\n - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat user-supplied data as untrusted or to prevent it from influencing the shell command parameters.\n
- Capability inventory: The skill utilizes the
BashandWritetools to perform setup and configuration tasks.\n - Sanitization: While
BUNDLERandPKG_MANAGERare validated usingcasestatements, theAGENTSvariable is used unsanitized in a shell context.
Recommendations
- AI detected serious security threats
Audit Metadata