gstack
An agent skill by garrytan, from garrytan/gstack. Tags: automation, multi-agent, router.
What it does
Router for the gstack skill suite. (gstack)
Install
With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:
npx skills add garrytan/gstack --skill gstack
Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):
git clone --depth 1 https://github.com/garrytan/gstack
cp -r gstack ~/.claude/skills/gstack
Safety box score
Grade F, high risk: 72 of 100 risk points, rated 2026-10-05. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.
gstack is a router that delegates into the rest of the gstack skill suite and runs local helper binaries for startup, telemetry, and logging. The main risks are broad autonomous routing into powerful workflows, executable prompts from tool output, and hidden behavior in helper scripts and telemetry plumbing.
| Category | Level | Why |
|---|---|---|
| Code & shell execution | 3 (high) | It runs local binaries on every session, e.g. `~/.claude/skills/gstack/bin/gstack-skill-start`, `.../gstack-learnings-log`, `.../gstack-skill-end`, and `.../gstack-telemetry-log` in `SKILL.md`. |
| Network access | 2 (moderate) | The suite ships telemetry plumbing with a remote Supabase endpoint in `supabase/config.sh` (`GSTACK_SUPABASE_URL=...supabase.co`) and the router records route telemetry. |
| Filesystem reach | 2 (moderate) | The workflow writes outside the project tree, including `~/.gstack/`, `~/.claude/skills/gstack`, and `$GSTACK_STATE_ROOT/analytics/`, plus it can ask the user to run upgrade/setup helpers. |
| Secrets & credentials | 1 (low) | This router does not directly read passwords or tokens, but it participates in telemetry and auth-adjacent workflows and tracks session identifiers like `SESSION_ID` and `TEL_START`. |
| Privilege & persistence | 1 (low) | It does not require sudo or system services, but it can steer the user toward install/upgrade operations that modify per-user tool state under home-directory paths. |
| Instruction hijack surface | 3 (high) | `SKILL.md` tells the agent to treat the skill file as executable instructions and to follow `GSTACK_INSTRUCTION_BEGIN` blocks from tool output, which is a prompt-injection surface. |
| Irreversible actions | 3 (high) | The router explicitly auto-invokes other skills when they match, including `/qa`, `/browse`, `/ship`, and `/land-and-deploy`, and says `When in doubt, invoke the skill`. |
| Transparency | 3 (high) | Key behavior is delegated to unread helper binaries under `~/.claude/skills/gstack/bin/`, and `extension/manifest.json` contains a long opaque `key` blob. |
Before installing
- Inspect `~/.claude/skills/gstack/bin/gstack-skill-start` and `.../gstack-skill-end` in the installed copy.
- Decide whether you want auto-routing to `/ship` and `/land-and-deploy`; if not, set `gstack-config set proactive false`.
- Check where `$GSTACK_STATE_ROOT` points and whether you are okay with analytics writes under that directory.
- Verify the telemetry endpoint/config in `supabase/config.sh` before letting the suite run.
Rated 2026-10-05 by an automated review of 207 files from the skill's repository.
Source
- Repository
- garrytan/gstack (all skills from this repository)
- Path
- SKILL.md
- Branch
- main
- Updated
- 2026-10-05
Related skills
- agents-build — Use to extend an existing agent project with memory, app integration, VPC, multi-agent, migration, model, browser, code interpreter, payments.
- app-platform-sandbox — Create and manage isolated container sandboxes for AI agent code execution.
- continual-learning — Orchestrate continual learning by delegating transcript mining and AGENTS.md updates to `agents-memory-updater`.
- coordinating-multi-space-devops-agent — Coordinate the AWS DevOps Agent across multiple AgentSpaces from one Claude Code session — route questions to the right space (prod vs staging vs knowledge).
- orchestrate — Use only when the user explicitly types `/orchestrate <goal>` to decompose a large task.
- ralph-loop — Start a Ralph Loop for iterative self-referential development.