The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to clone a repository from https://github.com/msitarzewski/agency-agents.git and run local shell scripts ./scripts/convert.sh and ./scripts/install.sh. Running scripts from unverified external sources is a primary vector for remote code execution and system compromise.
[EXTERNAL_DOWNLOADS]: The skill relies on fetching core functionality and persona files from a personal GitHub account that is not recognized as a trusted organization or a resource of the skill author.
[COMMAND_EXECUTION]: The installation process includes several shell commands that modify the local filesystem, such as copying files into the sensitive ~/.claude/agents/ configuration directory and creating symlinks.
[INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting external markdown files and piping their contents into the system prompts of various AI agents. Ingestion points: Files located within the agency-agents/ directory. Boundary markers: Absent, meaning there is no clear distinction between instructions and potentially malicious data in the external files. Capability inventory: The skill possesses the ability to execute shell scripts and write to system directories. Sanitization: The skill lacks any mechanism for validating or sanitizing the content of the external prompt files before they are loaded into the agent context.

agency-agents-ai-specialists