The Agent Skills Directory

[DYNAMIC_EXECUTION]: The validation scripts utilize dynamic execution methods to run project-specific tasks.
The Bash script scripts/validate.sh uses the eval command to execute shell strings constructed from the project's environment.
The PowerShell script scripts/validate.ps1 uses the Invoke-Expression cmdlet to execute commands.
[COMMAND_EXECUTION]: The skill executes shell and PowerShell commands as part of its "Validation Chain." These commands are derived from files within the project directory being analyzed.
In validate.sh, commands for linting, testing, and building are extracted from the scripts section of package.json and executed.
In validate.ps1, similar logic is used to extract and execute commands via PowerShell.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and executes data from potentially untrusted project configuration files.
Ingestion points: The scripts read and parse package.json, pyproject.toml, requirements.txt, go.mod, and Cargo.toml from the target project directory.
Boundary markers: No boundary markers or "ignore instructions" warnings are present when processing these configuration files.
Capability inventory: The skill has the capability to execute arbitrary shell and PowerShell commands, as well as read files and write to the console.
Sanitization: The scripts do not perform validation or sanitization on the command strings retrieved from configuration files before passing them to the execution engine (eval or Invoke-Expression).

apex-tier3