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

Layerkit Source Edit Client

skill-hariharapanigrahy-layerkit-layerkit-source-edit-client · by hariharapanigrahy

Integrate vendors by editing the production datalayer directly; style-match; run the client package verification command.

No reviews yet
0 installs
0 views
view→install

Install

$ agentstack add skill-hariharapanigrahy-layerkit-layerkit-source-edit-client

✓ 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-hariharapanigrahy-layerkit-layerkit-source-edit-client)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
5d 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 Layerkit Source Edit Client? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

layerkit-source-edit-client

Modify the existing production datalayer and production source files (adapters, registry, router, tests) directly from evidence. Do not ask the CLI for plan files or source patches.

Optional project.json source-edit hints:

{
  "sourceEdit": {
    "moduleRoot": "apps/platform/integrations",
    "denyEdit": ["**/legacy/**"]
  }
}

Protocol

  1. Read memory/runbooks/surface-inventory.json (from surfaces step). Every language must end as updated (with paths) or residual (with reason) before mark-done — Layerkit enforces this.
  2. Read existing production integration code, interfaces, mappers, tests, and package style.
  3. Run layerkit-deletion-first: remove/update stale code/docs/tests before adding files.
  4. Inspect the existing interface/datalayer and implement in production paths for each inventory language (or mark residual).
  5. For every new file/function/export, list what it replaces; if it replaces nothing, justify why it must exist.
  6. Update surface-inventory.json statuses after each language is done; re-run source-edit iteration until no language is pending.
  7. Run the client package verification command, such as the package's build/test/coverage CI target; then:

``bash layerkit doctor ``

  1. For any strategic redirect or broad rewrite, define what must pass, the proof artifact, and the fallback before edits start; run a small proof step first: an eval, end-to-end QA check, contract-heal case, release checklist item, package-level fixture, or before/after acceptance test.
  2. Strengthen executable tests relative to implementation size. Source-edit work that changes behavior must cover the client-package edit path, mapping semantics, deletion-first behavior, and the relevant CI/eval gate.

Mapping Rules

  • Generic rename pattern (email is only an example): if the vendor renames any field (e.g. emailemail_id) and the client still exposes the same domain value (getEmail()), update the existing vendor setter from the existing client getter: payload.setEmailId(event.getEmail()). Same rule for any evidence-backed rename. Do not guess renames from names alone — require docs/OpenAPI/changelog evidence.
  • If the vendor removes a field, remove the stale mapping and update tests. Do not keep a wrapper for removed vendor behavior.
  • If the vendor adds a field and the client interface/datalayer already has equivalent data, map the existing client field into the new vendor shape.
  • If no client source represents the required data, treat it as an unsupported datalayer gap: leave a localized TODO and handoff note with the missing interface/datalayer gap.
  • If research proved zero production field drift, do not invent edits: attest residual-no-field-edit with residual justification. Do not bump apiVersion or SDK alone as a substitute for real source work.

New Vendor Integrations

Full integrate of a new vendor into a multi-vendor (or single-vendor) package is end-to-end production work on the existing path, not a side project:

  1. Inventory siblings first — name concrete adapter/mapper/registry/test paths already shipping (from discover + style profile).
  2. Follow the existing path — place the new vendor beside siblings (same module root), mirror naming/error handling/privacy hooks, and wire the existing registry/router.
  3. Clone sibling tests — copy the nearest sibling test layout and adapt only what vendor evidence requires; run the client package verification command.
  4. No parallel facade — do not invent libs/-wrapper or a second registry when src/vendors/* + registry already exists.
  5. Add a new file only when no existing file should own the vendor-specific behavior; document which sibling pattern it follows and why modifying an existing file was not enough.
  6. Keep shared routing/runtime decisions in the client package. Layerkit only supplies agent skills and exact artifact rails.

Proof before broad work: package build/test green for the new vendor path (or residual TODO with missing datalayer gap). Handoff lists sibling reference paths, new production paths, registry wire, and PR URL.

Forbidden

  • AI calls in production adapter send paths
  • Invented field maps
  • Parallel facade beside an existing one
  • Handing off while client package verification or doctor fails
  • New adapter abstraction without explaining what existing file/function/export cannot be changed
  • Treating generated plans, package metadata, or staging files as source edits
  • Pin-only / apiVersion alone presented as full integrate (not full integrate)
  • Claiming source-edit done without production paths or residual-no-field-edit
  • Listing paths that do not exist on disk under the package root

Success

  • [ ] Production files updated by the agent after inspecting existing code (list paths) or residual-no-field-edit attested
  • [ ] Deletion-first pass complete; new files/functions/exports list what they replace
  • [ ] Existing style/topology honored from source evidence
  • [ ] Client package build/test/coverage command green
  • [ ] Package verification green for primary intents
  • [ ] Strategic redirects have a passing proof step before broad edits
  • [ ] Outcome checkpoints are recorded with what must pass, proof artifact, and fallback
  • [ ] Tests cover changed mapping semantics and deletion-first behavior when those areas changed

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.