Source-first Skill and MCP workflow for URL-based website cloning, capture, rebuild, and fidelity verification.
18 tools with complete input schemas and descriptions, but significant quality gaps. Tool names follow verb_noun pattern (generate_embed, inspect_url, detect_runtime, etc.), which is good. However, descriptions are often generic and lack actionable context for LLM selection. Parameters have types and descriptions, but many lack constraints (enums, ranges, format specs). Output schemas are not documented in the visible code. Error handling and recovery guidance are absent. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. The server exposes complex domain logic (browser automation, HAR replay, job queues) but does not guide LLMs on when to use each tool or how to compose them.
Perform a low-risk routing and approval preflight audit without browser capture.
Build a rebuild scaffold from a capture bundle.
Build a reproduction bundle from a capture bundle.
Cancel a clone job by ID.
Capture a complete reference bundle including static inspection, runtime trace, and optional breakpoint variants.
Clone a reference URL with full capture, audit, and optional rebuild.
Detect runtime capabilities including Node.js, Playwright, browser availability, and supported session modes.
Output schemas not documented. Tools return complex objects (bundles, reports, job states) but LLMs cannot see what fields to expect. This forces agents to guess at response structure and breaks downstream tool chaining.
Descriptions lack actionable context. Most descriptions (50-55 chars) state WHAT the tool does but not WHEN to use it or how it differs from similar tools. E.g., 'audit_reference_url' vs 'capture_reference_bundle', when should an LLM pick one over the other?
No tool annotations despite clear risk classifications. Tools marked WRITE (trace_runtime_sources, capture_reference_bundle, clone_reference_url, etc.) should declare destructiveHint=true. Tools marked READ_ONLY should declare readOnlyHint=true. This guides LLM safety reasoning.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2026-07-28+ | v2 |
Discover embeddable candidate URLs from a reference URL.
Enqueue a clone job in the job queue.
Generate an embed snippet (iframe or JSX) for a given URL and framework.
Inspect a reference URL to extract metadata, frame policy, platform detection, and candidate URLs.
List clone jobs in the job queue, optionally filtered by state.
Load a specific clone job by ID.
Plan the reproduction path based on candidates, source signals, and site profile.
Replay HAR (HTTP Archive) requests and build a replay report.
Run a clone job from the queue, either a specific job or the next pending job.
Trace runtime sources and interactions from a URL using browser automation.
Verify fidelity between reference and candidate bundles.
Parameters lack constraints. 'timeout_seconds', 'wait_seconds', 'viewport_width/height' have no min/max bounds. LLMs can pass absurd values (timeout_seconds=999999, viewport_width=1). Add numeric ranges and enum constraints.
No error handling or recovery guidance. Tools accept complex inputs (URLs, file paths, job IDs) but provide no guidance on what to do if validation fails. E.g., if 'url' is invalid, should the LLM retry with a different URL or call inspect_url first?