Skip to content
Agentmart
Submit a tool

intent-layer

Scaffold or audit a repo's hierarchical AGENTS.md Intent Layer.

Skill by jeanchastel

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/jeanchastel-intent-layer

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Scaffold or audit a repo's hierarchical AGENTS.md Intent Layer. Use when initializing context infrastructure, adding AGENTS.md nodes to a complex subsystem, or auditing existing context files for size and coverage.

Use cases: testing, devops, security

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Shell scripts

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/intent-layer

What's inside

Skill name
intent-layer
Description
Scaffold or audit a repo's hierarchical AGENTS.md Intent Layer. Use when initializing context infrastructure, adding AGENTS.md nodes to a complex subsystem, or auditing existing context files for size and coverage.
License
MIT
Instructions
54 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts

Files (4)

  • SKILL.md
  • analyze-structure.sh · scripts/
  • detect-state.sh · scripts/
  • estimate-tokens.sh · scripts/

SKILL.md (first lines)

# Intent Layer

Build and maintain a hierarchy of `AGENTS.md` context nodes so agents navigate the
repo like a senior engineer. The normative rules live in the Intent Layer section of
`governance/root-agents.md` in the resolved Arkira plugin. Full rationale and examples:
`governance/intent-layer-standard.md`.

## Workflow

1. Detect state.
   `bash scripts/detect-state.sh <repo>` reports `none`, `partial`, or `complete` and
   lists existing context nodes.

2. Measure.
   `bash scripts/estimate-tokens.sh <repo>` ranks immediate subdirectories by token
   weight (chars / 4). `bash scripts/analyze-structure.sh <repo>` lists candidate
   subsystem directories with file counts.

3. Decide.
   `AGENTS.md` is the single shared root context. Add a child `AGENTS.md` only when a
   directory exceeds roughly 20k tokens, owns a distinct responsibility, or holds a
   cross-cutting concern (place that node at the lowest common ancestor). Do not add a
   node for every directory, a simple utility, or a test folder unless it is genuinely
   complex.

4. Execute.
   Each node opens with a READ-FIRST directive and stays under 4k tokens. `CLAUDE.md`
   and `CODEX.md` stay as role overlays at the root only; never place them below the
   root and never duplicate the normative content from `AGENTS.md`.
   In a repository with a tracked `.arkira/config.json`, the active host writes `AGENTS.md`
   directly by default. Optional delegation uses
   `bin/arkira run ai-engineering/runtime/role-run.sh <repo> executor code_editing --prompt-file <file>`.
   Direct work creates no Executor receipt.
   When an area produces or consumes a contract artifact (see
   `docs/document-contract.md`), record a short Document Contract line in its
   `AGENTS.md` node: what it reads, what it writes, and which skills sit upstream
   and downstream.

5. Maintain.
   When a subsystem grows or its responsibility shifts, re-run the measure step and
   split or update the nearest node. Document hidden contracts and invariants in the
   nearest ancestor node.

## Capture questions

When documenting existing code in a node, answer:

1. What does this area own, and what is out of scope?
2. What invariants must never be violated?
3. What repeatedly confuses a new engineer here?
4. What patterns must always be followed?
5. What documents does this area read and write, and which skills produce or consume them?

From jeanchastel/arkira, MIT.

More from jeanchastel/arkira

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 23, 2026 (6 days ago)
Last release
Oct 23, 2026 (6 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from jeanchastel

Cite this page

intent-layer skill. Agent Mart. https://agentmart.dev/skills/jeanchastel-intent-layer (last updated 2026-10-24)