- Home
- Skills
- Agents & Orchestration
- Mini-App Spec
Mini-App Spec
Generate executable HTML5 specifications that reconcile PRDs, Figma designs, and code into a single source of truth.
Free
Works with the AI Agents you already use
Mini-App Spec
Example session with this skill installed
Start a new spec for the 'Checkout' feature. Title it 'Secure Checkout Flow'. Use the current PRD in docs/prd.md and the Figma link for the UI layout.
- Read your context and instructions
- Compiled the mini-app spec
I have initialized the specification for the Secure Checkout Flow. I've registered the PRD and Figma link as sources and decomposed the requirement into three core vertical flows for alignment.
Specification Status: Secure Checkout Flow
| Flow ID | Description | Risk Tier | Core | Status |
|---|---|---|---|---|
flow-01 | Guest Checkout with Stripe Integration | High | Yes | pending |
flow-02 | Registered User Saved-Card Payment | Med | Yes | pending |
flow-03 | Address Validation and Tax Calculation | Med | No | pending |
Source Mapping
src-01:docs/prd.md(Version: v2.1) - Mapped to all flows.src-02:figma://file/checkout-v4- Mapped to UI layout forflow-01andflow-02.
Next steps
- Review the decomposed flow list above and confirm the risk tiering.
- I will then execute
save --gate gate1to anchor these requirements in git. - We will begin per-flow alignment starting with
flow-01.
Connects securely to your tools. The creator never sees your data.
About this skill
Software requirements often drift from implementation, leading to broken acceptance criteria and "stale" documentation. This skill enforces a Single Source of Truth (SSOT) by generating a self-contained, executable HTML5 specification that acts as a machine contract, operable prototype, and BDD test suite. It bridges the gap between Figma designs and code, ensuring every feature is verified against a frozen behavioral fingerprint before acceptance.
What it does
- Executable Contract Generation produces a
mini-app-spec.htmlcontaining state machines, BDD scenarios, and embedded prototypes. - Automated Verification Gate freezes commands and collects evidence per scenario to prevent hand-waived passing status.
- Multi-Source Reconciliation maps requirements from PRDs, Figma (via MCP), and code into unified vertical user flows.
- Interactive Review Sessions launches a local browser session where users annotate the UI directly to generate structured change tickets.
- Mechanical Rule Enforcement blocks acceptance unless specific preconditions, like zero proposed annotations and valid source batches, are met.
How it works
- Resume or Initialize: Use
statusto recover context ornewto create a fresh artifact from the stateless engine template. - Decompose and Map: Register sources and decompose requirements into tiered flows, followed by a mandatory confirmation gate.
- Align and Prototype: Fill in states and transitions, then
openthe artifact for visual alignment and fingerprint freezing. - Reconcile and Accept: Run
record-runto collect evidence, thenacceptwith verbatim user confirmation to stamp the final gate.
Frameworks & tools
The system uses a Node.js CLI spec.mjs for all contract writes and a stateless HTML5/CSS/JS engine. It integrates with Git for replay anchors and recovery, supports Figma via MCP, and enforces CSP-compliant, zero-dependency templates.
Why this beats prompting it yourself
Generic prompts fail to maintain state across long implementation cycles and allow "status cheating" where AI claims success without evidence. This skill uses a mechanically enforced validator and git-anchored replay anchors to ensure the spec and implementation never diverge.
Use cases
- Documenting complex multi-state UI components that require high-fidelity prototyping.
- Auditing legacy codebases by reconciling existing behavior against new BDD requirements.
- Managing stakeholder sign-off with an operable, annotated "live" specification.
- Generating verifiable evidence for compliance-heavy features where manual testing is insufficient.
Known limitations
Requires a Git repository environment to function. The engine forbids external CDNs or heavyweight frameworks; all visual interactions must be self-contained within the HTML artifact.
Requires a local Git repository for replay anchors. Execution is restricted to Node.js environments. External CDNs and heavyweight JS frameworks are forbidden in templates.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
2 uses
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
Requires