Awesome Goal Prompts
A catalogue of over 300 ready /goal contracts for Claude Code, Codex and Cursor plus a skill that writes such contracts.
- 53 ★ on the repository
- Last commit
- CC0-1.0
Minimal access
Why this level
- Contracts are text for an agent: they run nothing themselves, but the agent will execute the verification commands they describe.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add prompts/awesome-goal-promptsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- Contracts are text for an agent: they run nothing themselves, but the agent will execute the verification commands they describe.
This is third-party code. Review the repository files before installing.
What it does
When an agent starts making broad edits, a contract helps: what to inspect, what not to touch, how to verify the result and when to stop. The repository holds over 300 such `/goal` contracts for engineering work. Each has `GOAL`, `CONTEXT`, `CONSTRAINTS`, `DONE WHEN`, `VERIFY`, `OUTPUT` and `STOP RULES` blocks. The main catalogue holds 159 source-backed examples from official docs, public GitHub threads, tutorials, forums and tool READMEs. Another 200 reusable patterns sit separately in seed-patterns so provenance is not diluted. Start with ten rescue contracts: fixing auth tests and lint, a green CI, a database migration with proof, a checkout crash investigation and others. A contract is copied whole and pasted to the agent. The `make-goal` skill in `skills/` works in create, tighten, review and catalogue-entry modes.
Who it is for. Developers whose agent drifts on a task and for whom "fix auth" is too vague.
Good fit when
- Auth tests and lint are red and a plain "fix it" invites the agent to change the public API.
- CI fails on tests, lint and type checks at once and you need a done criterion from local and remote runs.
- A database migration needs proof: application in a dev database and a schema comparison.
- Your own goal draft must be tightened or checked for gaps before the agent runs.
Not a fit when
- The task is small and a seven-block contract costs more than the edit itself.
- You need a ready automation with no human involved: contracts are text to paste, not a runner.
- A contract needs your project context and you cannot supply logs and paths.
Example request
Write a `/goal` contract for my task: auth tests and lint are red, the public API must not change, completion is proven by the output of passing commands.Limitations
Contracts must be adapted to your project: substitute your paths, verification commands and constraints. Example sources are public and the catalogue keeps the address, source type and an evidence phrase, but agent behaviour depends on the model and tool version, so no result is guaranteed. A contract demands proof of completion, such as passing command output, and the agent may fail to produce it: the stop rules then tell it to report and stop editing. The CC0 license lifts copying restrictions. There is no install command: contracts are copied from the README and the skill goes into the agent skills directory.
How to disable. Nothing is installed until you copy a contract. Delete the `make-goal` skill from the agent skills directory. Just stop pasting contracts and nothing else remains in the project.
FAQ
How do I use the catalogue?
Open the searchable catalogue on GitHub Pages or pick a contract from the README. Copy the whole block from `GOAL` to `STOP RULES`, substitute your details and paste it to the agent.
What does a contract contain?
Seven blocks: the goal (`GOAL`), context to inspect (`CONTEXT`), hard constraints (`CONSTRAINTS`), a done criterion (`DONE WHEN`), fresh verification (`VERIFY`), an output format (`OUTPUT`) and stop rules for uncertainty or risk (`STOP RULES`).
How do source-backed examples differ from seed patterns?
Examples have a public source address, a type and an evidence phrase. Seed patterns live in `prompts/seed-patterns.md`: reusable shapes without provenance that need adapting.
What does the make-goal skill do?
It creates, tightens or reviews `/goal` contracts and drafts a catalogue entry. If you ask both to write a goal and run it, it first produces the contract and waits for confirmation.
Can I add my own goal?
Yes. Find a thin category in the catalogue, draft the contract with the `make-goal` skill and add an entry following `CONTRIBUTING.md`: edit `data/source/entries.toml`, regenerate and run the validation commands.
Discussion
No questions or reviews yet.
Related
PromptsA skill that builds a Codex /goal command from a five-section template so a long task does not drift and passes the audit
A CLI compiler that takes rules, skills, agents and hooks written once and lays them out as native files for Claude Code, Cursor, Copilot and 47 more tools
Claude Code Prompts (RepoWise)
claude-code-prompts
Independently written prompt templates for AI coding agents: system prompts, tool prompts, delegation, memory and multi-agent coordination
A hundred ready loops for Claude Code, Cursor and Codex: the agent acts, runs a check and fixes the cause until the goal is met.
The author's personal Claude Code toolbox: meta-prompt and todo commands, skills to create skills, hooks and MCP servers, auditors and a Ralph loop
A skill that turns a vague request into a Goal Prompt with success criteria and stop conditions, plus a Loop Prompt if needed.
README in short
The README presents the catalogue as a set of contracts for rescuing a drifting agent. It covers a one-minute quick start, ten main contracts, sources, how to write a good goal, catalogue checks in CI and how to contribute an entry.