The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes a script (scripts/check-update.mjs) that performs network requests to the author's GitHub repositories (eliasstravik/gtm-skills and eliasstravik/gtm-agent) to retrieve version metadata and package information. This process is used to notify the user of available updates.
[COMMAND_EXECUTION]: The skill relies on several system-level commands to perform its core functions:
git is used for workspace version control, including initializing repositories and pushing/pulling changes.
gh (GitHub CLI) is used to create and share private repositories.
npx is used to run the installer for the skill package when updates are requested by the user.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from external websites, which presents a surface for indirect prompt injection:
Ingestion points: Facts are collected from "the company's own public site and other safe public sources" to fill in organization records (ORG.md).
Boundary markers: Ingested data is structured within Markdown templates (templates/ORG.md) to maintain the workspace schema.
Capability inventory: The agent has the ability to write files to the local ~/.gtm/ directory, execute shell commands (git, gh, npx), and perform network checks.
Sanitization: The skill procedure enforces a rule to mark unconfirmed or missing information as Unknown rather than inferring or inventing data, providing a check against potential injection from external content.

gtm-workspace