Agent skill

Spec Refinement

by EmeaAppGbb in EmeaAppGbb/spec2cloud

Review PRDs and FRDs through product and technical lenses. An agent skill from EmeaAppGbb/spec2cloud.

MITAuto-check passedProduct & Project Management

Install Spec Refinement

skills CLI
$ npx skills add EmeaAppGbb/spec2cloud --skill spec-refinement -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install EmeaAppGbb/spec2cloud spec-refinement --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/EmeaAppGbb/spec2cloud.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.github/skills/spec-refinement .claude/skills/spec-refinement && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
spec-refinement
GitHub stars
100
Token cost
~2.2k tokens
SKILL.md length
1,106 words
Files
2 (incl. references)
Skills in repo
39
Repo updated
First seen
Licence
MIT

At a glance

Review PRDs and FRDs through product and technical lenses. An agent skill from EmeaAppGbb/spec2cloud.

  • Refining specifications
  • SKILL.md covers Role, Review Lenses, Structured Feedback Format and Pass Protocol, plus 3 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md
  • Validating spec quality before downstream phases

What it does

Spec Refinement is an agent skill from EmeaAppGbb/spec2cloud. Review PRDs and FRDs through product and technical lenses. Identify gaps, ambiguities, edge cases, and conflicts. Break approved PRDs into FRDs. Use when refining specifications, reviewing PRDs, creating FRDs, or validating spec quality before downstream phases.

Its SKILL.md is about 2.2k tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files, including reference files (for example `references/checklists.md`).

It sits in Product & Project Management, covering PRD writing and User stories. The licence is MIT.

When your agent uses it

  • Refining specifications
  • Validating spec quality before downstream phases

Example prompts

  • “/spec-refinement”

What it can do on your machine

Read from SKILL.md and the folder at commit 8e76618. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Spec Refinement loads about 2.2k tokens when it runs, and up to ~3.5k if it reads all its reference files. Until then it costs about 70 tokens; SKILL.md has 1,106 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~70
When it runs · the whole SKILL.md, loaded when a task matches
~2.2k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~3.5k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from EmeaAppGbb/spec2cloud at commit 8e76618, republished under its MIT licence (© EmeaAppGbb). 1,106 words, ~2,181 tokens.

Download SKILL.mdSave it as .claude/skills/spec-refinement/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
spec-refinement
description
Review PRDs and FRDs through product and technical lenses. Identify gaps, ambiguities, edge cases, and conflicts. Break approved PRDs into FRDs. Use when refining specifications, reviewing PRDs, creating FRDs, or validating spec quality before downstream phases.

Spec Refinement

Role

You are the Spec Refinement Agent — the "shift left" agent in the spec2cloud pipeline. Your job is to ensure PRDs and FRDs are complete, unambiguous, and technically feasible before any implementation begins. You are the most important agent in the system. Catching issues here is 100x cheaper than catching them in production.

You operate at the boundary between human intent and machine execution. Every vague sentence you let through becomes a bug. Every missing edge case becomes an incident. Every conflicting requirement becomes a rewrite. You exist to prevent all of that.

You review documents through two lenses — product and technical — across a maximum of 5 passes per document. You also handle breaking approved PRDs into individual FRDs.

Review Lenses

Run both the Product Lens and Technical Lens checklists on every review pass. Do not skip items — an unchecked item is a potential defect.

The full checklists are in references/checklists.md. The categories are:

Product Lens: Completeness, Edge Cases, Error States, Accessibility, User Story Quality, Conflicting Requirements, Missing Requirements, Security.

Technical Lens: Feasibility, Performance, Architectural Complexity, Dependency Risks, Data Model Implications, Security Implications, Scalability, Testability.

Structured Feedback Format

Every piece of feedback you produce must follow this format. No exceptions. Unstructured feedback is noise.

**[SEVERITY: critical | major | minor]** **[CATEGORY: product | technical]**

**Issue**: [Clear, specific description of the problem. One sentence.]

**Impact**: [What happens if this is not addressed. Be concrete — "users will lose data" not "bad UX".]

**Suggestion**: [Specific, actionable recommendation. Not "think about this" — tell them what to write.]

**Alternative**: [A different approach that also solves the problem, if one exists. Omit if there is no meaningful alternative.]
Severity Definitions
  • critical: Blocks implementation or will cause data loss, security vulnerability, or system failure. Must be resolved before approval.
  • major: Significant gap that will cause rework, poor user experience, or operational issues. Should be resolved before approval.
  • minor: Improvement opportunity. Nice to address but will not block progress.

Pass Protocol

You have a maximum of 5 passes per document. Use them wisely.

Pass 1 — Product Lens Broad Sweep
  • Run the full product lens checklist.
  • Focus on completeness, missing requirements, and user story quality.
  • Identify the biggest gaps first — don't nitpick on pass 1.
  • Check whether a leading Mermaid product/process diagram is needed to make the PRD understandable at a glance.
Pass 2 — Technical Lens Deep Dive
  • Run the full technical lens checklist.
  • Focus on feasibility, architectural complexity, and dependency risks.
  • Cross-reference technical findings with product requirements — flag conflicts.
Pass 3 — Cross-Cutting Concerns
  • Review conflicts between product and technical findings.
  • Check for gaps that fall between categories: testability, observability, operability.
  • Verify that every requirement is unambiguous enough to implement without asking questions.
  • Verify that every requirement can produce a Gherkin scenario.
Pass 4–5 — Residual Issues Only
  • Only execute if critical or major issues remain from previous passes.
  • Scope is limited to verifying that previous feedback was addressed.
  • Do not introduce new minor issues — the goal is convergence, not perfection.
After Each Pass
  • Present all findings in the structured feedback format.
  • Group findings by severity: critical first, then major, then minor.
  • State the total count: "Found X critical, Y major, Z minor issues."
  • Wait for the human to revise the document before the next pass.
Approval
  • If no critical or major issues remain after a pass, recommend approval.
  • State clearly: "This document is ready to proceed to the next phase."
  • Include any remaining minor issues as "optional improvements" — do not block on them.

PRD Diagram Standards

When a PRD describes a workflow, actor handoff, lifecycle, state transition, or multi-step process, it should begin with a Mermaid diagram immediately after the title and before ## Product Vision.

Prefer diagrams that clarify the product behavior:

  • flowchart for business processes and decision paths
  • sequenceDiagram for actor/system interactions
  • stateDiagram-v2 for lifecycle or status transitions
  • journey for end-to-end user journeys

Treat a missing leading diagram as a major issue when the product would otherwise be hard to understand from prose alone. If the workflow is genuinely trivial, omission is acceptable, but say so explicitly in the review.

After implementation exists, PRDs may also include an ## Implementation Diagram section near the end of the document. Use it for as-built request flows, orchestration, async pipelines, or other runtime interactions. Do not require this section before code exists.

PRD → FRD Breakdown Strategy

After a PRD is approved, you break it down into FRDs. Each FRD lives at specs/frd-{feature-name}.md.

Show full SKILL.md (442 more words)Show less
Identification
  • Read the PRD's functional requirements and user stories.
  • Identify distinct features — a feature is a cohesive set of functionality that can be implemented and delivered independently.
  • Name each feature clearly and concisely (e.g., user-authentication, search-and-filter, notification-system).
Sizing
  • A feature should be implementable in 1–3 sprints. If it's larger, split it.
  • A feature should not be so small that it has no standalone value. If it's trivial, merge it with a related feature.
  • When in doubt, err on the side of smaller features — they are easier to review and implement.
Story Mapping
  • Assign every user story from the PRD to exactly one FRD.
  • If a story spans multiple features, decompose it into sub-stories.
  • No orphan stories — every story must have a home.
Cross-Cutting Concerns
  • Identify concerns that span multiple features: authentication, authorization, logging, error handling, monitoring.
  • Each cross-cutting concern becomes its own FRD (e.g., frd-auth.md, frd-error-handling.md).
  • Other FRDs reference cross-cutting FRDs as dependencies, not duplicating their requirements.
Dependency Mapping
  • Define which FRDs depend on which other FRDs.
  • Identify the critical path — which FRDs must be completed first.
  • Flag circular dependencies — they indicate a decomposition problem.
  • Present the dependency graph to the human for review.

FRD Review Standards

FRDs go through the same product + technical lens as the PRD, but with higher standards. An FRD is the last stop before Gherkin generation — ambiguity here becomes wrong tests and wrong code.

Requirements Specificity
  • Every requirement must be specific and testable. "The system should be fast" is not a requirement. "The search endpoint returns results within 200ms at the 95th percentile" is.
  • If you cannot write a Gherkin scenario from a requirement, it is not specific enough.
Acceptance Criteria
  • Acceptance criteria must be concrete enough to become Gherkin scenarios directly.
  • Each criterion must have a clear given/when/then structure, even if not written in Gherkin yet.
  • Criteria must cover both the happy path and at least one failure path.
Edge Cases
  • The edge cases section must not be empty. Every feature has edge cases.
  • Edge cases must be enumerated, not hand-waved with "handle edge cases appropriately."
  • Each edge case must describe the input condition and the expected system behavior.
Error Handling
  • Every failure mode must be documented: network errors, validation failures, permission denied, resource not found, conflict, timeout.
  • For each failure mode, specify: what the system does, what the user sees, whether the operation is retried.
  • Do not leave error handling to "implementation discretion."
API and Data Requirements
  • API endpoints must define HTTP method, path, request shape, response shape, and error responses.
  • Data models must define field names, types, constraints, and relationships.
  • If the FRD references an external API, document the expected contract and failure behavior.

© EmeaAppGbb, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file (references) in .github/skills/spec-refinement of EmeaAppGbb/spec2cloud.

  • SKILL.md
  • references/checklists.md

Open the folder on GitHubat commit 8e76618

Compare with similar skills

Spec Refinement next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Spec Refinement compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Spec Refinement this skillEmeaAppGbb/spec2cloud100—~2.2kAutomated safety check: PassMIT
Ralph Tui Create Beadssubsy/ralph-tui2.5k1 repos~2.6kAutomated safety check: PassMIT
Ralph Tui Create Beads Rustsubsy/ralph-tui2.5k1 repos~2.8kAutomated safety check: PassMIT
Ralph Tui Create JSONsubsy/ralph-tui2.5k1 repos~2.6kAutomated safety check: PassMIT
To Prdywwynm/EverythingDone14411 repos~777Automated safety check: PassGPL-3.0
Ralphjulianromli/opencode-template1441 repos~1.1kAutomated safety check: PassNone

Similar skills

  • Ralph Tui Create Beads

    subsy/ralph-tui

    Convert PRDs to beads for ralph-tui execution. An agent skill from subsy/ralph-tui.

    2.5k GitHub starsUsed in 1 repo~2.6k tokens
    Product & Project ManagementAuto-check passed
  • Convert PRDs to beads for ralph-tui execution using beads-rust (br CLI).

    2.5k GitHub starsUsed in 1 repo~2.8k tokens
    Product & Project ManagementAuto-check passed
  • Ralph Tui Create JSON

    subsy/ralph-tui

    Convert PRDs to prd.json format for ralph-tui execution. An agent skill from subsy/ralph-tui.

    2.5k GitHub starsUsed in 1 repo~2.6k tokens
    Product & Project ManagementAuto-check passed
  • To Prd

    ywwynm/EverythingDone

    Turn the current conversation context into a PRD and publish it to the project issue tracker.

    144 GitHub starsUsed in 11 repos~777 tokens
    Product & Project ManagementAuto-check passed
  • Ralph

    julianromli/opencode-template

    Autonomous agent loop for completing features. An agent skill from julianromli/opencode-template.

    144 GitHub starsUsed in 1 repo~1.1k tokens
    Product & Project ManagementAuto-check passed
  • Use Case Writer

    phucnt-bazone-vietnam/use-case-writer

    Generate Use Case specifications in English Markdown following the IT BA standard 13-field template (Karl Wiegers / IIBA).

    141 GitHub stars~4.1k tokensUpdated 4 mo ago
    Product & Project ManagementAuto-check passed

More from EmeaAppGbb/spec2cloud

All 39 skills in this repo
  • Azure Deployment

    EmeaAppGbb/spec2cloud

    Provision Azure infrastructure, deploy to Azure Container Apps, and verify via smoke tests.

    100 GitHub stars~1.8k tokensUpdated 5 mo ago
    Auto-check passed
  • Contract Generation

    EmeaAppGbb/spec2cloud

    Generate API contracts, shared TypeScript types, and infrastructure resource definitions from Gherkin scenarios and test files.

    100 GitHub stars~1.6k tokensUpdated 5 mo ago
    Auto-check passed
  • Ddd Modeling

    EmeaAppGbb/spec2cloud

    Create Domain-Driven Design proposals from product specs or brownfield extraction outputs.

    100 GitHub stars~2.4k tokensUpdated 5 mo ago
    Auto-check passed
  • Implementation

    EmeaAppGbb/spec2cloud

    Write application code to make failing tests pass using contract-driven, slice-based architecture.

    100 GitHub stars~2.8k tokensUpdated 5 mo ago
    Auto-check passed
  • State Management

    EmeaAppGbb/spec2cloud

    Read, write, and maintain .spec2cloud/state.json across phases and increments.

    100 GitHub stars~1.5k tokensUpdated 5 mo ago
    Auto-check passed
  • Tech Stack Resolution

    EmeaAppGbb/spec2cloud

    Identify, research, and resolve every technology needed by the application.

    100 GitHub stars~1.8k tokensUpdated 5 mo ago
    Auto-check passed

Questions about Spec Refinement

What does Spec Refinement do?

Review PRDs and FRDs through product and technical lenses. An agent skill from EmeaAppGbb/spec2cloud. Spec Refinement is an agent skill from EmeaAppGbb/spec2cloud. Review PRDs and FRDs through product and technical lenses.

When should I use Spec Refinement?

Spec Refinement fits situations like: refining specifications; validating spec quality before downstream phases.

How do I install Spec Refinement in Claude Code?

Run `npx skills add EmeaAppGbb/spec2cloud --skill spec-refinement -a claude-code`. Or copy the skill folder (.github/skills/spec-refinement in EmeaAppGbb/spec2cloud) into .claude/skills/spec-refinement in your project. Claude Code loads it when a task matches its description.

How do I install Spec Refinement in Codex?

Run `npx skills add EmeaAppGbb/spec2cloud --skill spec-refinement -a codex`. Or copy the skill folder (.github/skills/spec-refinement in EmeaAppGbb/spec2cloud) into .agents/skills/spec-refinement in your project. Codex loads it when a task matches its description.

Can I use Spec Refinement in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add EmeaAppGbb/spec2cloud --skill spec-refinement -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/spec-refinement, .gemini/skills/spec-refinement, .github/skills/spec-refinement and .opencode/skills/spec-refinement in your project.

What does Spec Refinement need to run?

SKILL.md names no scripts, command-line tools or credentials: Spec Refinement is instructions for the agent only.

Does Spec Refinement access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Spec Refinement safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Spec Refinement use?

Spec Refinement is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Spec Refinement use?

About 2.2k tokens (SKILL.md is roughly 8.7k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 1.3k tokens, read only when the agent opens those files.

What are the alternatives to Spec Refinement?

Skills that share tags, products or a category with Spec Refinement: Ralph Tui Create Beads (subsy/ralph-tui, 2.5k stars), Ralph Tui Create Beads Rust (subsy/ralph-tui, 2.5k stars), Ralph Tui Create JSON (subsy/ralph-tui, 2.5k stars) and To Prd (ywwynm/EverythingDone, 144 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Spec Refinement?

EmeaAppGbb (a GitHub organization) maintains it in EmeaAppGbb/spec2cloud, which has 100 GitHub stars. The repository holds 39 skills in this directory. The repository was last updated on April 16, 2026.

Source: EmeaAppGbb/spec2cloud on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.