The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/absorb-agents-skills.sh is susceptible to command injection. It constructs a PowerShell command string by interpolating directory names from the .agents/skills folder. A malicious directory name containing shell metacharacters, such as '; calc #, would result in the execution of arbitrary PowerShell commands when the synchronization script is run.\n- [PRIVILEGE_ESCALATION]: The skill instructs the agent to execute PowerShell scripts using the -ExecutionPolicy Bypass flag. This action explicitly circumvents system security policies designed to prevent the execution of untrusted or unsigned scripts on the host Windows machine.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the local filesystem to drive command-line operations without sufficient sanitization.\n
Ingestion points: Folder names in the .agents/skills directory processed by scripts/absorb-agents-skills.sh.\n
Boundary markers: Absent; there are no checks to validate or delimit folder names before processing.\n
Capability inventory: File deletion (rm -rf), recursive copying, junction creation, and arbitrary command execution via powershell.exe in scripts/absorb-agents-skills.sh and scripts/restore-windows.ps1.\n
Sanitization: Absent; folder names are interpolated directly into shell strings without escaping or validation.\n- [COMMAND_EXECUTION]: The synchronization process involves recursive deletion (rm -rf) and the modification of filesystem junctions. If paths are manipulated via malicious folder names, this could lead to unintended data loss or the redirection of sensitive configuration directories.

sync-skills-wsl