ag

Warn

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the ag binary to perform recursive text searches across files and directories. The provided setup scripts (scripts/install.sh and scripts/install.ps1) execute shell commands and use sudo to acquire administrative privileges for installing the tool via system package managers including apt, dnf, pacman, and apk.
  • [EXTERNAL_DOWNLOADS]: The installation scripts facilitate the download of The Silver Searcher utility from official and well-known system package registries such as Homebrew for macOS, WinGet for Windows, and various Linux distribution repositories.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it searches and retrieves content from arbitrary and potentially untrusted files within the repository.
  • Ingestion points: The ag tool reads and outputs the content of files from the project directory and subdirectories (File: SKILL.md).
  • Boundary markers: Absent; there are no instructions or delimiters provided to the agent to ignore or isolate instructions that may be embedded within the searched files.
  • Capability inventory: The skill allows deep filesystem searching; the resulting context is then processed by the agent, which may have additional capabilities depending on the environment.
  • Sanitization: Absent; the content of searched files is returned as raw text without any filtering or sanitization.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 8, 2026, 04:18 PM
Security Audit — agent-trust-hub — ag