dna init

Make a project agent-ready: board + skill + AGENTS.md + git hooks.

One command bootstraps everything an AI coding agent needs to work DNA-style in this project:

 .dna// SDLC board (Genome via the kernel) /skills/dna-sdlc-cli/ the SDLC workflow skill (agentskills.io), projected per --tools (.claude/skills, .github/skills, .cursor/skills, ...) AGENTS.md the canonical instruction surface (agents.md/v1 — read by 28+ agent tools) git hooks Work-Item commit trailers

The assets ship inside dna-cli as an embedded onboarding scope of real Kinds and are materialized by the SDK's own byte-faithful readers/writers — one Kind, N regenerable projections. AGENTS.md serves every tool at once; Gemini CLI users can point GEMINI.md at it.

With --from, the skills + AGENTS.md come from a DISTRIBUTED onboarding pack (your team's own conventions) instead of the embedded scope. The pack is fetched and validated with the same machinery as dna install (untrusted-input defenses included) but only PROJECTED into tool directories — nothing from the pack is written to the .dna/ source. Review the projected files before committing: a skill is agent instructions; treat a third-party pack like a dependency.

Idempotent: re-running never overwrites an existing file unless --force is given; the summary reports what was created vs skipped.

Examples:

 dna init # here, board '-dev' dna init --scope acme-dev # explicit board scope dna init --tools all # every supported tool dir dna init --tools claude,cursor # explicit projection set dna init --dir ../other-project # initialize another directory dna init --from github:acme/onboarding-pack@v1 # your team's pack dna init --from local:../onboarding-pack # offline authoring loop

Generated from the command definitions

This page is introspected from the dna Click command tree by scripts/gen_cli_docs.py, so it stays in lockstep with dna init --help.

dna init

Make a project agent-ready: board + skill + AGENTS.md + git hooks.

One command bootstraps everything an AI coding agent needs to work DNA-style in this project:

 .dna// SDLC board (Genome via the kernel) /skills/dna-sdlc-cli/ the SDLC workflow skill (agentskills.io), projected per --tools (.claude/skills, .github/skills, .cursor/skills, ...) AGENTS.md the canonical instruction surface (agents.md/v1 — read by 28+ agent tools) git hooks Work-Item commit trailers

The assets ship inside dna-cli as an embedded onboarding scope of real Kinds and are materialized by the SDK's own byte-faithful readers/writers — one Kind, N regenerable projections. AGENTS.md serves every tool at once; Gemini CLI users can point GEMINI.md at it.

With --from, the skills + AGENTS.md come from a DISTRIBUTED onboarding pack (your team's own conventions) instead of the embedded scope. The pack is fetched and validated with the same machinery as dna install (untrusted-input defenses included) but only PROJECTED into tool directories — nothing from the pack is written to the .dna/ source. Review the projected files before committing: a skill is agent instructions; treat a third-party pack like a dependency.

Idempotent: re-running never overwrites an existing file unless --force is given; the summary reports what was created vs skipped.

Examples:

 dna init # here, board '-dev' dna init --scope acme-dev # explicit board scope dna init --tools all # every supported tool dir dna init --tools claude,cursor # explicit projection set dna init --dir ../other-project # initialize another directory dna init --from github:acme/onboarding-pack@v1 # your team's pack dna init --from local:../onboarding-pack # offline authoring loop

dna init [OPTIONS]

Options

Option	Description
--dir	Project directory to initialize (default: current directory). (default: .)
--force	Overwrite existing onboarding files (skill projections, AGENTS.md). The board Genome is never rewritten — an existing board is verified and kept.
--from	Project a DISTRIBUTED onboarding pack instead of the embedded one: github:owner/repo[/subdir][@ref], local:<path>, or a bare directory path. The pack must carry at least one agentskills.io Skill bundle; a root AGENTS.md replaces the embedded instruction surface (absent: the embedded one is used, with a note). Pack content is validated with the same defenses as dna install and only PROJECTED into tool directories — combine with dna install <same-uri> when you also want it on the board.
--help	Show this message and exit.
--json	Machine-readable summary.
--scope	Board scope name (default: '-dev', slugified — the pilot convention for a dev-time SDLC board).
--tools	Comma-separated agent tools to project the SDLC skill for (claude, copilot, cursor, opencode — or 'all'). The SKILL.md format is identical across tools; only the directory differs. (default: claude,copilot)