bp
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local applications and system tools such as
open,xdg-open,osascript, andpkillbased on user-driven tasks. It implements internal validation logic withinscripts/bp.shto ensure that only GUI applications registered in standard system directories (e.g.,/Applicationsor/usr/share/applications) can be launched. It explicitly refuses to execute arbitrary shell commands likermorshby checking them against the list of discovered desktop applications.\n- [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it ingests untrusted data from the local environment, including application names from.desktopfiles and file names from the user's home directory (e.g.,~/Downloads). Maliciously crafted file names could potentially influence the agent's behavior when it processes the output of the shortcut tool.\n - Ingestion points: The
scripts/bp.shscript reads.desktopfiles to discover applications and usesfindandstatto list files in the user'sDownloads,Desktop, andDocumentsfolders.\n - Boundary markers: The instructions do not specify explicit boundary markers for data returned by the shell script to the agent.\n
- Capability inventory: The skill has capabilities for file system searching (
find), file writing (appending to the TSV store and creating files in~/Documents/bp), and process/command execution (launch,stop).\n - Sanitization:
scripts/bp.shvalidates aliases, URL schemes, and target existence. It filters out system and shell processes from itsstop(quit) command to prevent accidental system disruption.
Audit Metadata