A small, portable skill pack for moving from a product goal to verified implementation.
It is intentionally generic: no application code, customer data, internal URLs, project history, or company-specific operating rules are included.
/authoring-goal-docs— research the codebase and write a durable product goal./grill-for-unknowns— challenge assumptions, dependencies, and proof before planning./gap-analysis— turn the agreed goal into a prioritized task queue./gap-close— execute the approved queue through/orchestrating-subtasks.
Supporting skills:
/efficient-fable— reserve Fable for judgment and delegate bounded work./calling-claude— run Claude/Fable non-interactively when available./codex— run resumable Codex CLI sessions and request an independent review or second opinion./implement— apply the shared preflight and verification loop.
Copy the skills you want into Codex's skills directory:
git clone https://github.com/hsuperman/goal-driven-coding-workflow.git
mkdir -p ~/.codex/skills
cp -R goal-driven-coding-workflow/skills/<skill-name> ~/.codex/skills/Repeat the copy command for each skill you want. Start with the four workflow skills and add the supporting skills required by your chosen model setup.
- A Codex-compatible skill host.
- Git and a local repository to inspect.
- Codex CLI for
/codex. - Claude CLI and authentication only if using
/calling-claudeor/efficient-fablewith Claude/Fable.
The pack describes a workflow; it does not install models, authenticate providers, or make deployments automatic. Review each skill and adapt it to your own repository's safety rules before giving agents write access.
This is a small, rewritten workflow pack. It does not contain application code or a verbatim copy of any upstream skill.
/grill-for-unknownsis adapted from Nico Bailon's publicgrill-for-unknownsrepository. That project credits Matt Pocock'sgrill-with-docs,grilling, anddomain-modeling, together with Thariq's “A Field Guide to Fable: Finding Your Unknowns”./codexis a compact adaptation of the publiccodexskill from gstack, by Garry Tan. The gstack project is MIT licensed; this pack does not include its runtime, telemetry, or other gstack files./efficient-fableis adapted from Builder.io's publicefficient-fableskill./calling-claude,/implement, and the goal/gap/orchestration skills are local workflow guidance written for this pack. No gstack provenance was found for them.
The pack is not affiliated with gstack, Matt Pocock, Thariq, Anthropic, or OpenAI.