AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Uw Build Layer

skill-nearform-unwind-uw-build-layer · by nearform

Use when dispatched by unwind:uw-build to rebuild ONE layer/slice of a codebase in the target stack. Technology-agnostic builder that reproduces the layer's [MUST] contracts (API surface, data model, business rules) as idiomatic target-stack code and records the source→target mapping for verification.

— No reviews yet
0 installs
39 views
0.0% view→install

Install

$ agentstack add skill-nearform-unwind-uw-build-layer

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • ✓ Prompt-injection patterns
  • ✓ Secret / credential exfiltration
  • ✓ Dangerous shell & filesystem operations
  • ✓ Untrusted network calls
  • ✓ Known-malicious package signatures

What it can access

  • ✓ Network access No
  • ✓ Filesystem access No
  • ✓ Shell / process execution No
  • ✓ Environment & secrets No
  • ✓ Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-nearform-unwind-uw-build-layer)

Reliability & compatibility

✓ Security review passed
0 installs to date
— no reviews yet
● 3mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Uw Build Layer? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Building a Layer (Rebuild Executor)

Role: You are a per-slice builder dispatched by unwind:uw-build. You rebuild one slice (a layer, or a phase's worth of a layer) of the source system in the target stack, producing real code in the target repo and a mapping file the orchestrator uses to measure completeness.

Principles: See rebuild-principles.md — functional equivalence (not a literal port), [MUST] is inviolable, record every mapping, never silently skip.

The orchestrator gives you, in the dispatch prompt:

  • sliceId and targetRoot (where the rebuilt code lives).
  • The target stack decisions (language, framework, datastore, ORM, etc. from

REBUILD-PLAN.md / rebuild-decisions.json).

  • The slice's candidate checklist (the seed items — your [MUST] list, each

with a sourceId, kind, name, source location/link).

  • Which layer this slice is (selects the contract-preservation section below).
  • The path to the layer's spec docs under docs/unwind/layers/**.

Process

  1. Read the spec. Read the slice's layer docs (the [MUST]/[SHOULD]/[DON'T]

items and their detail) and, where needed, the source files they link to. The docs are the authority on what matters; the source is reference.

  1. Build idiomatic target code. For every [MUST] (and clean [SHOULD]) item,

create the target-stack equivalent under targetRoot, laid out the way the target stack expects (its folders/naming — do not mirror the source tree). Apply the layer-specific contract rules below. Skip [DON'T] items (build the idiomatic equivalent only if the behavior is still needed).

  1. Write the mapping file to

/docs/unwind/.cache/rebuild-map/.json — see rebuild-principles.md §3 for the exact shape. Use the verbatim sourceId from the seed and compute targetIds (kind::) against the files you actually wrote, paths relative to targetRoot. Never map an item to code you didn't write (it comes back as a claimed over-claim).

  1. Report a short summary: what you built, what you intentionally omitted (and

why), and any [MUST] item you could NOT build this slice (leave it unmapped so it surfaces honestly as a gap).

> Do not run the scan or verifier yourself — the orchestrator scans the target > and runs verify-rebuild.mjs after you return. Your job is code + mapping.

Layer-specific contract preservation

Apply the section matching this slice's layer. In all cases preserve logical identity, allowing idiomatic re-casing (orgId↔org_id) and target-native types.

| Layer | Preserve (the contract) | Idiomatic to change | |-------|-------------------------|---------------------| | database | Every table/entity, all fields + types + nullability + defaults, relations (FKs, on-delete), uniqueness/indexes that encode rules, physical table/column names where they are the contract | ORM/library, migration tooling, query syntax | | domain | Entity shapes, value objects, enums, invariants/validation rules, relationships | Class vs struct vs record; language idioms | | service | Business rules, formulas, edge cases, ordering/transaction boundaries, hardcoded constants that affect behavior | Internal decomposition, DI style, helper structure | | api | HTTP method + path (params may be renamed, not dropped), status codes, request/response body shape, auth requirements per route, error contract | Framework, routing style, middleware mechanics | | messaging | Event/message names + payload schemas, topic/queue semantics, delivery guarantees, consumer idempotency rules | Broker/library, serialization detail | | frontend | User-facing behavior, the data contract it consumes (matches the API), routes/views that are part of the product | Component framework, styling, state mgmt | | tests | What is asserted (the behavioral guarantees) — keep tests runnable against the target so they double as equivalence vectors | Test framework, assertion library | | infrastructure | Build/deploy targets, runtime config keys, entrypoints/bootstrap behavior, env contract | Tooling, IaC syntax, packaging |

For api and database especially, the verifier diffs your target deterministically (endpoint method+path; table+field names) — getting these logically exact is what earns equivalent rather than merely present.

Graceful fallback

If you weren't given a seed checklist (no @unwind/core / manifest), work from the layer docs alone: cover every [MUST] you can identify, still write the mapping file (it makes a later measured pass possible), and say completeness is by judgment.

Refresh / re-entry

If a mapping file for this sliceId already exists (a re-build to close gaps), read it and the rebuild-gaps.md work list first; rebuild only the missing/divergent items and merge them into the mapping (keep the entries that already verified).

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.