The Agent Skills Directory

[PROMPT_INJECTION] (LOW): The groom.sh script is susceptible to indirect prompt injection because it parses untrusted data from backlog files to perform filesystem operations.\n
Ingestion points: groom.sh processes all .md files in the backlog/ directory and backlog/ROADMAP.md.\n
Boundary markers: No delimiters or warnings are used to prevent the script from processing maliciously crafted content as legitimate instructions.\n
Capability inventory: The script executes git log and git branch and performs filesystem existence checks ([[ -e \"$filepath\" ]]). This enables an attacker to confirm whether specific files exist on the host by placing those paths in the 'Files to create' section of a backlog file.\n
Sanitization: There is no validation or sanitization of the file paths or keywords extracted from the backlog files before they are used in commands.

backlog