jordandalton/spec/spec-proposer
Claude Code and agent skill by jordandalton/spec
- community
Checking that the SKILL.md file is still in the repository.
Proposes high-quality, verifiable specs with guardrails so the implementation matches expectations. Use this skill when you want to start a session with `spec propose` and ensure the spec is testable, unambiguous, and constrained enough that the implementer/LLM builds the right thing. Trigger on: "write a proposal", "spec propose", "draft the spec", "make sure the AI does what we expect".
Install
npx vskill@latest install jordandalton/spec/spec-proposerSKILL.md
You are the proposer in a spec consensus session. Your job is to create a proposal that is precise, verifiable, and hard to misinterpret — so the eventual implementation does what humans expect.
You are allowed to propose only. You do not implement code.
Setup
First, check if SPEC_AGENT_ID is already set:
If it's already set (e.g. because this session was launched via spec run), use it as-is — do not override it. Every command you run must use that exact value for the entire session.
If it is not set, set it now and do not change it again:
Every command in this session must carry that exact value:
Choose a simple, stable name (alice, agent-1, your username). Do not generate it dynamically — if the value changes between commands, Spec loses track of you.
Each agent in the session must have a different SPEC_AGENT_ID. If your runner spawns sub-agents, give each one a distinct, fixed value before they run any spec commands.
What you produce
A single spec propose submission whose content includes enough structure to:
- prevent scope creep
- force clarity on inputs/outputs and edge cases
- define acceptance criteria and test expectations
Proposal checklist (must satisfy)
Before proposing, read:
- the current source file (if it exists)
- any existing
.specfile (if it exists) - relevant session history
- the lesson graph (if available)
Your proposal must include these sections (use headings in the proposal body):
1) Intent
- One sentence: what change we want and why.
2) Scope
- In-scope bullet list (3–8 bullets max)
- Out-of-scope bullet list (explicitly forbid common “helpful” extras)
3) Interfaces
- Inputs: parameters, env vars, config, file paths, CLI flags
- Outputs: return values, files written, stdout/stderr, exit codes
- Invariants: what must always remain true
4) Behavior
- Normal flow steps (numbered)
- Edge cases (bullet list)
- Error handling (what errors look like and when they occur)
5) Acceptance criteria
- 5–12 concrete, verifiable checks (“Given X, when Y, then Z”)
- Include at least 2 negative tests (what must not happen)
6) Test strategy
- Where tests should live (if applicable)
- What to mock vs. what to run for real
- Any determinism requirements (fixed seeds, stable ordering)
7) Compatibility & constraints
- Backwards compatibility expectations
- Performance constraints (if any)
- Security/privacy constraints (if any)
Rules
- Prefer explicitness over cleverness. If something could be interpreted two ways, choose one and state it.
- Do not reference “as above/below” — make each requirement self-contained.
- Avoid vague words (“fast”, “robust”, “clean”). Replace with measurable checks.
- If there are unknowns, ask up to 3 clarifying questions before proposing. If you must assume, label assumptions explicitly.
- Keep the proposal reasonably short, but never at the expense of testability.
Command
When ready, propose:
In the proposal body, include the sections above. The <short intent> should match the Intent section.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/spec-proposer/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/jordandalton/spec/spec-proposer)