| name | layer-new |
| description | Scaffold a new layer in a project's docs/layers/ tree following the feature-layer architecture (principle 28). A layer is a bounded concern (security, data, ui, infrastructure, domain) with its own invariants, decisions, gotchas, patterns, and feature narratives. Use when: "create a new layer", "add security layer", "scaffold layer", "start tracking <concern> separately", "/layer-new", "add bounded concern". Operates on the kb-skeleton structure; idempotent -- will not overwrite existing layers.
|
| user-invocable | true |
| model | sonnet |
/layer-new -- scaffold a project layer
Creates docs/layers/<layer-name>/ with the full template structure
defined by principle 28.
When to use
- Starting to track a new bounded concern in a long-running project
- Refactoring sprawling cross-cutting code into a documented layer
- Onboarding a new team member who needs the layer map
When NOT to use
- One-off scripts or pet projects with <5 features (overhead not
justified)
- Layer name describes a directory (
src is not a layer)
- The "layer" is actually one feature in disguise -- use
/feature-new
inside an existing layer instead
Arguments
/layer-new <layer-name> [--purpose "..."] [--principles P-NN,P-MM]
<layer-name> -- kebab-case, single word preferred. Examples:
security, data, image-processing, observability.
--purpose -- one-sentence purpose. If omitted, prompt the user.
--principles -- comma-separated Tier 1 principle IDs that govern
this layer (e.g. P-02,P-21). If omitted, leave placeholder in
README.
Direction (what to do, in order)
Step 1 -- Verify environment
Check the current working directory:
- Is it a git repository? Run
git rev-parse --show-toplevel. If
not, ask the user whether to initialize one (offer git init +
private GitHub repo per global rule).
- Does
docs/ exist? If not, create it.
- Does
docs/layers/README.md exist? If not, copy from
<claude-code-skills-checkout>/templates/kb-skeleton/docs/layers/README.md.
- Check if
docs/layers/<layer-name>/ already exists. If yes,
stop with a message -- do not overwrite. Suggest
/feature-new <layer> <slug> instead.
Step 2 -- Validate layer name
- Must be lowercase kebab-case (
[a-z][a-z0-9-]*).
- Must not start with
_ (reserved for templates).
- Must not be a generic file-system name (
src, tests, docs,
build).
- If invalid, refuse with a clear message and a suggested fix.
Step 3 -- Copy the template
Source: <claude-code-skills-checkout>/templates/kb-skeleton/docs/layers/_LAYER-TEMPLATE/
Destination: <repo>/docs/layers/<layer-name>/
Copy the entire directory tree. Preserve subdirectory structure (kb/
and features/). Result:
docs/layers/<layer-name>/
├── README.md
├── history.md
├── kb/
│ ├── invariants.md
│ ├── decisions.md
│ ├── gotchas.md
│ └── patterns.md
└── features/
└── _FEATURE-TEMPLATE.md
Step 4 -- Fill placeholders
In every file under the new layer, replace:
<layer-name> -> the actual layer name
<Layer name> -> Title Case of the layer name (e.g. "Security",
"Image Processing")
In README.md specifically:
**Purpose:** <one sentence...> -> the --purpose argument value,
or prompt the user
## Governing principles list -> populate from --principles arg,
or leave the placeholder bullets in place for the user to fill
In history.md:
- Insert a "Layer created" entry at the top with today's date
(YYYY-MM-DD) and the originating reason. Prompt the user for the
reason if not provided.
Step 5 -- Register the layer
Update docs/layers/README.md:
- Add a row to the
## Layer index table:
| <layer-name> | <purpose> | active |
- If a cross-layer Mermaid graph exists, add a node for the new layer
with no edges (user will add edges as dependencies form).
Step 6 -- Wire to project state
If the project has feature_list.json at repo root, leave it alone --
features get added by /feature-new. Do not edit feature_list.json
from this skill.
If the project has AGENTS.md, suggest (but do not auto-edit) adding
the new layer to the "Source-of-truth docs" table if multiple layers
exist.
Step 7 -- Confirm and suggest next step
Print a summary:
Layer created: docs/layers/<layer-name>/
Files: 1 README, 1 history, 4 kb/, 1 feature template
Suggested next steps:
1. Fill governing principles in docs/layers/<layer-name>/README.md
2. Write the first feature: /feature-new <layer-name> <slug>
3. Add the first invariant when it earns its place
Blueprints (files this skill writes from)
templates/kb-skeleton/docs/layers/_LAYER-TEMPLATE/ -- the source
tree to copy
templates/kb-skeleton/docs/layers/README.md -- the layers index
template (used only if missing)
Gotchas
- Renaming a layer is not idempotent. If the user runs
/layer-new wrong-name then realizes they wanted right-name,
manually rename the directory and update references. This skill
does NOT detect or fix duplicates.
- Layer name collision with existing directories. If
docs/<layer-name>/ exists at the docs/ root (not under
docs/layers/), refuse and ask the user which they want -- there
is no automatic merge.
- Templates may have moved. If the template path
<claude-code-skills-checkout>/templates/kb-skeleton/docs/layers/_LAYER-TEMPLATE/
does not exist (e.g. user is on a different machine), fall back to
reading from
https://github.com/AnastasiyaW/claude-code-config/tree/main/templates/kb-skeleton/docs/layers/_LAYER-TEMPLATE
via gh api.
- Encoding boundary. When writing files with Cyrillic content,
follow the global rule from
~/.claude/rules/api-utf8-posting.md --
always specify encoding="utf-8" explicitly.
Troubleshooting
| Symptom | Cause | Fix |
|---|
| "Layer already exists" | Directory docs/layers/<name>/ present | Use /feature-new to add to it, or pick a different name |
| Template files missing | Path moved or different machine | Pull templates from public repo via gh api repos/AnastasiyaW/claude-code-config/contents/templates/kb-skeleton/docs/layers/_LAYER-TEMPLATE |
| Layers README not updated | docs/layers/README.md had no ## Layer index table | Open file manually, add table per the kb-skeleton template |
| Validator warns about layer | validate_kb_links.py flagged broken link | Layer is fine; broken link is in a feature doc inside it. Run python scripts/build_kb_graph.py from the project root for the full health report. |
Implementation note
The bulk of the work is file copy + placeholder replacement. No
dynamic logic is needed; the template files do all the structural
heavy lifting. Keep this skill deterministic and idempotent -- it
must be safe to invoke twice on the same layer (second call should be
a no-op with a clear message).