The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/install_all.py script utilizes subprocess.run to execute npx skills add, which handles the installation of member skills. This functionality is controlled via a dry-run default setting, and the SKILL.md instructions explicitly state that commands should only be executed after the user agrees.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of ecosystem members from the vendor's repository at github.com/fatih-developer/fth-skills. This is an expected management function for a skill ecosystem and targets the vendor's verified namespace.
[INDIRECT_PROMPT_INJECTION]: As an orchestrator, this skill ingests artifacts produced by other tools. This ingestion surface is documented in references/handoff-contract.md, which provides mitigation by instructing the agent to treat all incoming artifacts as data to be verified rather than ground truth, specifically requiring re-validation before any destructive actions are taken.

ecosystem-security