goal-brief
Draft a /goal contract body (Objective, Read first, Constraints, Validate, Document, Checkpoints, Stop when) from your stated intent, ready to paste after /goal. Always includes anti-reward-hacking and anti-scope-creep clauses.
- License 许可
- MIT
- 最近更新
- 2026-11-15
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- goal-brief
- description
- Draft a /goal contract body (Objective, Read first, Constraints, Validate, Document, Checkpoints, Stop when) from your stated intent, ready to paste after /goal. Always includes anti-reward-hacking and anti-scope-creep clauses.
- allowed-tools
- Read, Grep, Bash
- argument-hint
- <what you want the agent to accomplish>
- disable-model-invocation
- true
02 资源文件 2
- 许可证LICENSE1.0 KB
- 参考文档UPSTREAM.md1.5 KB
03 SKILL.md 正文
Goal brief
<!-- Template adapted from davidondrej/skills goal-loop (template only; every
product/feature claim in that source -- lifecycle states, pause/resume,
create_goal, subscription auth -- is fabricated and discarded; LICENSE +
UPSTREAM.md in this directory carry attribution). Our real /goal <text>
is a Claude Code built-in: it installs a session-scoped Stop hook that
blocks stopping until the condition holds and auto-clears on success
(see memory: goal-command-sets-session-scoped-stop-hook). -->
Draft the contract body below from $ARGUMENTS (or the conversation's stated
intent if no argument was given). Emit only the contract body -- no
/goal prefix, no explanation before or after it. The user pastes /goal
in front of it themselves.
Before drafting
- Identify the actual files/dirs the agent should read first -- list what
really exists in this repo (
ls,grep, or what's already open in context), never an invented path. - Identify the real validation command for this project (
pnpm test,pytest -q, etc. -- checkpackage.json/pyproject.toml, don't guess). - One objective, one stop condition -- this is a contract, not a backlog.
If the objective needs more than ~4,000 characters, put the detail in a
file (
PLAN.md) and point the goal at it instead of inlining it. - Use literal strings for paths, commands, and issue numbers -- exact, not paraphrased.
Template
**Objective:** <one-sentence objective>
**Read first:** <files/dirs/PLAN.md/issue -- literal paths>
**Constraints:** <what must NOT change: public API, files, libs, conventions>. Do not refactor unrelated code. Do not add dependencies.
**Validate:** `<exact command>` after each change
**Document:** write concise, targeted documentation for the changes -- new .md files or focused updates to existing docs
**Checkpoints:** work in checkpoints; log progress briefly
**Stop when:** <verifiable condition>, OR when further changes require human/product input. Do not delete, skip, weaken, or narrow tests to make the goal pass.
The two appended sentences -- scope-creep in Constraints, reward-hacking in Stop when -- are non-negotiable. Include them verbatim on every contract, even when the user's stated intent doesn't mention them.
Examples
Migration:
**Objective:** Migrate this project from Pydantic v1 to v2.
**Read first:** pyproject.toml, src/, tests/
**Constraints:** no public API changes; keep imports backwards-compatible via shims if needed; no new dependencies. Do not refactor unrelated code.
**Validate:** `pytest -q` after each change
**Document:** update the migration notes in docs/ as shims or behavior changes land
**Checkpoints:** work in checkpoints; log progress briefly
**Stop when:** full suite passes with zero deprecation warnings, OR when a change requires architecture decisions. Do not delete, skip, weaken, or narrow tests to make the goal pass.
Coverage lift:
**Objective:** Raise coverage in src/auth/ from ~38% to >=75%.
**Read first:** src/auth/, tests/auth/, AGENTS.md
**Constraints:** no new deps; mirror existing test style; do not modify production code unless strictly required for testability. Do not refactor unrelated code.
**Validate:** `pytest --cov=src/auth --cov-report=term-missing`
**Document:** note any testability seams added, in the module's README or docstrings
**Checkpoints:** work in checkpoints; log coverage delta each one
**Stop when:** coverage >=75% AND all tests pass, OR when uncovered code needs design changes. Do not delete, skip, weaken, or narrow tests to make the goal pass.
04 同仓库更多 Skills
review-adversarial-first
Operator-invoked adversarial review: 3 refute-by-default lenses plus a deterministic judge against a PR, branch, or path, then a decision on whether /ce-code-review's topic fan-out runs. Off the default merge path since the 2026-10-17 merge-and-end ruling; run only when asked for by name.
memory-mine
Run one deterministic, budgeted directed-capture batch for an explicit Claude Code project store. Use manual workspace mode only when cross-agent cass sessions are intentionally in scope.
papers
Search preprints from arXiv, bioRxiv, and medRxiv without API keys. Supports search, lookup, extraction, citations, enrichment, and paper downloads that bundle PDFs plus linked GitHub code.
papers
Search preprints from arXiv, bioRxiv, and medRxiv without API keys. Supports search, lookup, extraction, citations, enrichment, and paper downloads that bundle PDFs plus linked GitHub code.
wsl-cdp
Drive the user's real, logged-in Windows browser (Chrome, Edge, or Brave) from a Claude Code session running inside WSL2, over the Chrome DevTools Protocol - read authenticated dashboards and consoles, take real screenshots, open and read pages with the user's sessions. Use when a task needs the user's existing web logins and the session runs in WSL2, where the Claude in Chrome extension cannot work. Not for native-Windows or macOS Claude Code (use the official Claude in Chrome extension), not for anonymous scraping or CI (use headless Playwright inside WSL), not for getting past login walls (those are handed to the human).
token-eater
[BETA] Spend a subscription's idle, about-to-expire credits on safe, machine-verified cleanup. You pick the service (Grok / Codex / Claude) and a token-heavy maintenance skill (de-monolithize, de-slop, simplify, dead-code); that service runs the whole skill -> gate -> self-review -> draft-PR loop on its own credits, in an isolated worktree. token-eater verifies the gate and never merges. Manual invocation only during beta.