The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous shell commands for auditing environment configurations and creating safe shell scripts. The recommended patterns, such as set -euo pipefail and proper variable quoting, align with established security best practices for shell environments.
[EXTERNAL_DOWNLOADS]: The skill instructions involve downloading and installing popular open-source CLI tools from well-known package managers (Homebrew, APT, Cargo) and GitHub repositories (e.g., tmux-plugins/tpm). These are standard development workflows and do not involve untrusted or suspicious sources.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its primary function of auditing user-provided files. * Ingestion points: External data enters the context when the agent audits shell configurations or reviews shell scripts as instructed in guides/01-shell-audit.md and guides/03-shell-scripting.md. * Boundary markers: The instructions lack explicit delimiters to isolate untrusted user data from the agent's execution instructions. * Capability inventory: The agent is tasked with generating audit reports and suggesting code fixes based on static analysis; it does not invoke or execute the content of the audited files. * Sanitization: There is no mention of sanitizing or escaping the content of the files being reviewed.

terminal-bash-weapon