The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple local binaries and shell commands for configuration, session tracking, and environment detection, including gstack-config, gstack-update-check, and gstack-timeline-log. It also leverages gh and glab CLI tools for interaction with git hosting platforms.
[REMOTE_CODE_EXECUTION]: The skill performs dynamic execution of shell code using eval on the output of gstack-slug and source on the output of gstack-repo-mode. Additionally, it performs git fetch and merge operations on an artifacts repository using a remote URL defined in local configuration files.
[DATA_EXFILTRATION]: The skill collects telemetry data (skill name, duration, outcome) and transmits it via the gstack-telemetry-log binary. It also reads the agent's platform configuration file ~/.claude.json to identify MCP server settings.
[PROMPT_INJECTION]: The skill ingests untrusted data from various project documentation files (e.g., README.md, ARCHITECTURE.md) and pull request descriptions. While it has access to high-privilege tools like Bash and Write, it lacks defined boundary markers or sanitization routines to isolate this external content from the agent's instruction context.

document-release