Target Support

aspens separates two concepts:

• Target — where generated docs are written and how the AI tool discovers them
• Backend — which CLI generates the content

aspens supports three targets: Claude, Codex, and OpenCode.

Targets

Claude

Claude support is hook-aware and file-aware.

Published artifacts:

Artifact	Path
Instructions	CLAUDE.md
Skills	.claude/skills/<name>/skill.md
Hooks	.claude/hooks/...
Settings	.claude/settings.json

Claude reads CLAUDE.md on startup and auto-activates skills through hooks and activation rules.

Codex

Codex support is document-driven rather than hook-driven.

Published artifacts:

Artifact	Path
Instructions	AGENTS.md
Skills	.agents/skills/<name>/SKILL.md
Directory scope	<dir>/AGENTS.md

Codex reads the root AGENTS.md and can use directory-scoped AGENTS.md files for local context.

OpenCode

OpenCode support (added in 0.9.0) is document-driven, like Codex, but reuses Claude’s skills layout.

Published artifacts:

Artifact	Path
Instructions	AGENTS.md
Skills	.claude/skills/<name>/skill.md

OpenCode reads the root AGENTS.md and the .claude/skills directory. The backend drives opencode run --format json.

Picking a target

Each target is explicit — pass exactly one:

npx aspens doc init --target claude
npx aspens doc init --target codex
npx aspens doc init --target opencode

Caution

--target all was removed in 0.9.0. With three targets — and codex and opencode both owning AGENTS.md — all was ambiguous. Run doc init --target <one> once per target instead. Passing --target all now errors with the valid list.

Adding a second target to an existing repo

You don’t need to start from scratch. aspens reuses existing repo understanding:

# Add Codex output to a repo that already has Claude docs
npx aspens doc init --target codex

aspens uses existing CLAUDE.md and .claude/skills as context when generating Codex or OpenCode output.

Note

codex and opencode both write AGENTS.md, so they can’t be configured as targets in the same repo. aspens rejects that combination with a clear error — pick one document-driven target per repo.

Target vs Backend

The target and backend are independent. You can mix and match:

# Claude format, generated by Claude CLI
npx aspens doc init --target claude --backend claude

# Codex format, generated by Codex CLI
npx aspens doc init --target codex --backend codex

# Codex format, generated by Claude CLI
npx aspens doc init --target codex --backend claude

Configuration

Target and backend settings persist in .aspens.json:

{
  "targets": ["claude"],
  "backend": "claude"
}

Once configured, commands like doc sync and doc impact use the saved settings automatically.

Sync behavior

The git post-commit hook works for all configured targets:

npx aspens doc sync --install-hook

One hook, all targets updated. Claude-specific activation hooks remain Claude-only.