The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute local binaries located in ~/.claude/skills/gstack/bin/. These commands are used for configuration management (gstack-config), update checks (gstack-update-check), and data logging (gstack-timeline-log).
[REMOTE_CODE_EXECUTION]: The preamble and various commands utilize dynamic execution patterns, including eval "$(...)" and source <(...) targeting local gstack binaries. Additionally, the /learn stats command uses bun -e to execute a JavaScript string for processing JSONL data. These patterns are used to dynamically load environment variables and process local data files.
[DATA_EXFILTRATION]: The skill contains logic for telemetry collection and artifact synchronization. Telemetry (usage events and metadata) is sent via gstack-telemetry-log, and artifact syncing (sharing designs/reports) is performed via a remote GitHub repository. Both features are governed by user-prompted configuration settings during the initial setup flow.
[PROMPT_INJECTION]: The skill ingests untrusted data from learnings.jsonl, which contains insights and patterns captured in previous sessions. While this creates a surface for indirect prompt injection, the data is primarily used for informational displays or processed via specific scripts rather than being directly interpolated into high-privilege system instructions without oversight.

learn