The Agent Skills Directory

[DYNAMIC_EXECUTION]: The preamble performs dynamic code execution by using eval on the output of ~/.claude/skills/gstack/bin/gstack-slug and source on the output of ~/.claude/skills/gstack/bin/gstack-repo-mode. This allows the skill to execute shell commands generated at runtime by these local utilities.
[COMMAND_EXECUTION]: The skill relies extensively on a set of local binaries located in ~/.claude/skills/gstack/bin/. These are executed during the preamble and at various stages of the retrospective workflow. Examples include gstack-config, gstack-update-check, gstack-telemetry-log, and gstack-learnings-search. It also uses the open command to launch external URLs in the browser.
[EXTERNAL_DOWNLOADS]: The skill performs update checks via gstack-update-check and includes functionality for an "inline upgrade flow". It also features an "Artifacts Sync" mechanism that can publish project artifacts to a private GitHub repository using gstack-brain-sync.
[DATA_EXFILTRATION]: Usage data, including skill identifiers and repo names, is logged to ~/.gstack/analytics/skill-usage.jsonl. The skill also uses gstack-telemetry-log to send event-type data, outcomes, and session IDs to a remote service, which represents a data exfiltration surface.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from git log output, CHANGELOG, TODOS.md, and Various session logs to generate its analytical narrative. It possesses a high attack surface as it lacks boundary markers or sanitization logic while maintaining access to powerful tools like Bash and Write.

retro