The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the $ARGUMENTS variable to set a threshold value in a shell script without sanitization. In Phase 1, the variable MIN_LINES is assigned from the first argument and then used in an echo statement: echo " Minimum duplication threshold: $MIN_LINES lines". This usage allows for command injection via subshell expansion if malicious input is provided as a skill argument.
[EXTERNAL_DOWNLOADS]: The skill performs runtime installation of the jscpd tool from the NPM registry using the command npm install -g jscpd. This installation is conducted without version pinning, which can lead to the execution of untrusted or modified code if the package is compromised in the registry.
[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary source code from the user's project directory, creating a vulnerability surface for indirect prompt injection.
Ingestion points: Local source directories (e.g., src/, lib/, app/) are read and processed by find, grep, and jscpd tools as seen in Phases 1 and 3.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between the skill's instructions and the content of the files being analyzed.
Capability inventory: The skill maintains capabilities for file system access, directory creation, shell command execution (jscpd, grep, jq), and package installation.
Sanitization: The skill does not perform validation or sanitization on the file content before it is passed to analysis tools or summarized in the final report.

duplication-detect