plan-ceo-review

An agent skill by garrytan, from garrytan/gstack. Tags: planning, product-strategy, review, strategy.

What it does

CEO/founder-mode plan review. (gstack)

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add garrytan/gstack --skill plan-ceo-review

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/plan-ceo-review ~/.claude/skills/plan-ceo-review

Safety box score

Grade F, high risk: 67 of 100 risk points, rated 2026-10-16. 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.

This skill is a high-privilege review workflow that runs many shell commands, reads and writes files under `~/.claude/` and the gstack state tree, and can send plan content to external services for research and outside review. The main risks are broad local filesystem reach, repeated command execution, and instruction-heavy text that pressures the agent to auto-decide and follow fetched content.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	3 (high)	It explicitly runs shell commands and hooks, including `~/.claude/skills/gstack/bin/gstack-skill-start`, `bash -c ... exec bash "$S"`, `gstack-skill-end`, and multiple `git`/`jq`/`codex` invocations.
Network access	2 (moderate)	It can send plan content to external systems via `WebSearch`, Aside, and the Codex outside-review path, so repository text leaves the machine during research and review.
Filesystem reach	3 (high)	It reads and writes outside the project in `~/.claude/` and `$GSTACK_STATE_ROOT`, including plan archives, review logs, analytics, tmp files, and cleanup like `rm -rf "$__OUTSIDE_TMP"`.
Secrets & credentials	2 (moderate)	It touches local gstack config and brain-cache state under `~/.claude/` and can surface cached project/user context; it does not explicitly read keychains or browser stores, but it does reach into local state that may be sensitive.
Privilege & persistence	1 (low)	It does not request `sudo`, services, or shell-startup edits, but it does create directories, set permissions, and persist state in user-home tooling paths.
Instruction hijack surface	3 (high)	The skill text repeatedly tells the agent to treat the file as executable instructions, auto-decide, and follow instruction blocks from tool output, e.g. `Treat the skill file as executable instructions` and `never ask confirmation`-style directives.
Irreversible actions	2 (moderate)	It can rewrite review artifacts, logs, plan files, and even promote docs/designs when allowed, but it is still a review workflow and does not itself deploy, pay, or post externally.
Transparency	0 (none)	The shipped files are plain text and readable; no binaries, base64 blobs, or downloaded installers are present in the reviewed skill files.

Before installing

Read `SKILL.md` and `sections/review-sections.md` for every path that writes under `~/.claude/` or `$GSTACK_STATE_ROOT`.
Run it once on a scratch repo with no sensitive plan text to see exactly what gets written to home-state and analytics.
Check whether you are comfortable sending plan content to external review/search services.
Confirm you want an agent that can execute shell commands and install hooks, not just generate prose reviews.

Rated 2026-10-16 by an automated review of 5 files from the skill's repository.

Source

Repository: garrytan/gstack (all skills from this repository)
Path: plan-ceo-review/SKILL.md
Branch: main
Updated: 2026-10-05

gstack-openclaw-ceo-review — Use when asked to review a plan, challenge a proposal, run a CEO review, poke holes in an approach, think bigger about scope.
launch — When the user wants to plan a product launch, feature announcement, or release strategy.
seo-plan — Strategic SEO planning for new or existing websites. Industry-specific templates, competitive analysis, content strategy, and implementation roadmap.
claude-md-progressive-disclosurer — Optimize, slim, or restructure CLAUDE.md/AGENTS.md with progressive disclosure and zero information loss. Use when the user explicitly asks to audit, 精简, 瘦身.
grill-with-docs — A relentless interview to sharpen a plan or design, which also creates docs (ADR's and glossary) as we go.
advisor — Advisor mode. Consult a stronger (or different) model at key checkpoints: before major decisions, when stuck on an error, and before declaring a task done.