dbs-install-skill
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell scripts (
install-skill.sh) and PowerShell scripts (windows-junction.ps1) to manage filesystem links. Evidence:skills/dbs-install-skill/scripts/install-skill.sh link <skill-name-or-path>inSKILL.md. - [COMMAND_EXECUTION]: On Windows, the skill bypasses execution policies to run a local helper script. Evidence:
powershell_args=(-NoLogo -NoProfile -NonInteractive -ExecutionPolicy Bypass -File "$helper" ...)inscripts/install-skill.sh. - [INDIRECT_PROMPT_INJECTION]: The skill processes external skill directories which could contain malicious
SKILL.mdfiles or frontmatter designed to influence the installer's behavior. Evidence:name="$(skill_name "$src")"andvalidate_interface_gate "$candidate"inscripts/install-skill.sh. - Ingestion points: Reads
SKILL.mdfiles from user-specified paths or local directories. - Boundary markers: Uses
---YAML delimiters andawkto extract specific metadata fields. - Capability inventory: Perform symbolic link creation (
ln -s), Junction creation (PowerShell), file removal (rm), and directory creation (mkdir) across~/.agents,~/.claude,~/.grok, and other agent homes. - Sanitization: Includes basic validation for illegal characters in skill names (
.|..|*/*) and checks if targets are real directories before overwriting.
Audit Metadata