container filesystem

Container filesystem layout

This page is a consolidated reference for every host-path → container-path mapping that harnessed creates when it launches a stack. Use it to understand where files land inside the container, diagnose mount problems, and author persist: / init: entries correctly.

Two invariants govern all mounts:

1. Profile is assembled ahead of time. Nothing in $XDG_DATA_HOME/harnessed/profiles/<stack>/ is resolved, templated, or patched at container start — what harnessed build wrote is what the container sees, byte-for-byte.
2. No secrets baked into images, and credentials never resolved by harnessed build. Every secret surface (.credentials.json, SSH agent socket, opt-in git credentials) reaches the container only at launch time, as a bind-mount or env-file, and only when the host-side file exists. harnessed build never invokes varlock and never touches any credential at all — see Secrets guide for the deliberate separation between build and launch.

---

Container identity

Value	Detail
Container user	harnessed (renamed from Ubuntu's default ubuntu)
$HOME	/home/harnessed
CONTAINER_HOME constant	paths.CONTAINER_HOME = Path("/home/harnessed")

The base image (catalog/base/Dockerfile.harnessed-base) sets ENV HOME=/home/harnessed. Every in-container path that starts with $HOME or ~ resolves there.

---

Profile mounts (read-only)

Assembled by harnessed build into $XDG_DATA_HOME/harnessed/profiles/<stack>/. All mounts below are read-only (ro). Only harnesses that actually consume each file receive the mount (claude/omp/opencode for the .claude/ tree; claude-only for .mcp.json).

Host path	Container path	Notes
profiles/<stack>/.mcp.json	$HOME/.mcp.json	claude only; passed to --mcp-config
profiles/<stack>/settings.json	$HOME/.claude/settings.json	claude, omp, opencode
profiles/<stack>/.claude/skills/	$HOME/.claude/skills/	claude, omp, opencode
profiles/<stack>/.claude/commands/	$HOME/.claude/commands/	claude, omp, opencode
profiles/<stack>/.claude/agents/	$HOME/.claude/agents/	claude, omp, opencode
profiles/<stack>/.claude/hooks/	$HOME/.claude/hooks/	claude, omp, opencode
profiles/<stack>/.claude/rules/	$HOME/.claude/rules/	claude, omp, opencode
profiles/<stack>/hatago.config.json	$HOME/hatago.config.json	all harnesses

where profiles/ expands to $XDG_DATA_HOME/harnessed/profiles/ (defaults to ~/.local/share/harnessed/profiles/).

---

Project workspace mount

The project directory (or a parent via --mount-folder) is mounted path-mirrored: the same absolute path on the host appears at the same absolute path inside the container. This means the agent sees host paths directly — no path translation needed.

Host path	Container path	Mode
<mount_path>	<mount_path> (same)	rw

The agent's working directory (cwd) is set to <project_path> (or --agent-start-folder if specified). With --mount-folder, <mount_path> is a parent of <project_path> — the whole subtree is visible but the agent starts inside the project.

---

Claude session history mounts (read-write)

These dirs are sourced from the host $HOME/.claude/ and mounted read-write so Claude's session history, file-tracking, and todos persist across container restarts. Dirs are created on the host automatically if they don't exist.

Host path	Container path	Mode
~/.claude/projects/	$HOME/.claude/projects/	rw
~/.claude/file-history/	$HOME/.claude/file-history/	rw
~/.claude/tasks/	$HOME/.claude/tasks/	rw
~/.claude/session-env/	$HOME/.claude/session-env/	rw
~/.claude/todos/	$HOME/.claude/todos/	rw

---

Credential mounts

Claude OAuth token

Host path	Container path	Mode	Notes
~/.claude/.credentials.json	$HOME/.claude/.credentials.json	ro	Only mounted when file exists
$XDG_STATE_HOME/harnessed/<instance>/claude.json	$HOME/.claude.json	rw	Per-instance onboarding stub; seeded from host ~/.claude.json (account identity only, never the OAuth token)

<instance> is harnessed-<stack>-<sha1[:8] of project path>.
$XDG_STATE_HOME defaults to ~/.local/state.

The stub at .claude.json marks hasCompletedOnboarding: true so Claude Code skips its first-run login screen. The real authentication is ~/.claude/.credentials.json (OAuth token) — the stub never contains token material.

omp (Oh My Pi)

Host path	Container path	Mode	Notes
~/.omp/agent/	$HOME/.omp/agent/	rw	omp harness only; full host-state sharing

SSH signing/auth agent + git config (always forwarded when present)

Forwarded automatically whenever the host agent socket is live — no forward_git_credentials needed. Safe as a default because the agent exposes no key material and gates every sign/auth behind a host-side 1Password approval or YubiKey touch, and the git config is a public signing-key reference, not a secret. All read-only unless noted; only mounted when the host file/socket exists.

Host path	Container path	Mode	Notes
~/.1password/agent.sock (Linux) or macOS equivalent	$HOME/.1password/agent.sock	socket	1Password SSH agent; sets $SSH_AUTH_SOCK
gpg-agent SSH socket (gpgconf --list-dirs agent-ssh-socket)	$HOME/.gnupg-sockets/S.gpg-agent.ssh	socket	YubiKey fallback; sets $SSH_AUTH_SOCK when 1Password absent
~/.config/git/ (or ~/.gitconfig)	$HOME/.config/git/ (or $HOME/.gitconfig)	ro	Carries user.signingkey, gpg.format=ssh, etc.

git credentials (opt-in, forward_git_credentials: true in stack.yaml)

The secret-bearing surface — only mounted when a stack sets forward_git_credentials: true, and only when the corresponding host file/dir/device exists. All read-only unless noted.

Host path	Container path	Mode	Notes
~/.config/gh/hosts.yml	$HOME/.config/gh/hosts.yml	ro	gh oauth token — so gh authenticates as you
~/.gnupg/pubring.kbx, trustdb.gpg, gpg.conf, gpg-agent.conf, sshcontrol	$HOME/.gnupg/<name>	ro	Non-secret GPG surface only; private-keys-v1.d/ is explicitly excluded
~/.ssh/config, known_hosts, *.pub	$HOME/.ssh/<name>	ro	Non-secret SSH surface
~/.ssh/<name> where <name> is in stack.yaml ssh_keys:	$HOME/.ssh/<name>	ro	Opt-in private keys; honored only from user overlay catalog
YubiKey USB device	--device passthrough	—	Linux only; when YubiKey is attached

---

Persist mounts (read-write)

Recipe persist: entries create bind-mounts that survive --fresh restarts. The scope × location axes determine how the host path is constructed. For the full authoring schema (field names, VCS control, valid combinations), see Recipe authoring — persist:.

scope: workspace, location: host — per-worktree, harnessed-managed

Host path template	Container path	Mode
persist/<recipe>/<sha1[:8] of realpath(project_path)>/<name>/	$HOME/<name>/	rw

persist/ expands to $XDG_DATA_HOME/harnessed/persist/ (defaults to ~/.local/share/harnessed/persist/). The key is the SHA-1[:8] of the resolved launch path — two worktrees of the same git checkout get separate dirs.

scope: project, location: host — cross-worktree, harnessed-managed

Host path template	Container path	Mode
persist/<recipe>/<sha1[:8] of git-common-dir>/<name>/	$HOME/<name>/	rw

The key is derived from git rev-parse --path-format=absolute --git-common-dir — the same path for every worktree of a given checkout, so all worktrees share one host dir. Falls back to the workspace hash (with a warning) when project_path is not inside a git repository. Does not survive an independent git clone of the same repo (git-common-dir changes).

scope: global, location: (not specified) — real host path, allowlisted

Host path	Container path	Mode
<path> (the real host path, allowlisted)	<path> (same, path-mirrored)	rw

The path is mounted path-preserving (host path == container path) so the tool finds its data where it expects. Gated by $XDG_CONFIG_HOME/harnessed/persist-allowlist (default ~/.config/harnessed/persist-allowlist). Several sensitive dirs (~/.ssh, ~/.aws, ~/.gnupg, ~/.config/harnessed, $HOME itself) are hard-denied regardless of the allowlist.

Path-preserving ≠ $HOME-relative. The pod's $HOME is /home/harnessed, so a host dotdir mounted this way lands at /home/you/.foo, not ~/.foo — a tool that looks under $HOME will not find it. The launcher exports HOST_HOME into the attach shell for exactly this case; point the tool at the mirrored path from the recipe's init:. The pulumi recipe is the reference implementation (see the Pulumi guide):

persist:
  - scope: global
    path: ~/.pulumi        # the real host dir: credentials.json + the plugin cache

init:
  run: export PULUMI_HOME="$HOST_HOME/.pulumi"

scope: workspace|project, location: in_repo — inside the workspace, no extra mount

No additional bind-mount: the entry lives inside the path-mirrored workspace mount, which is already rw. vcs: ignored causes harnessed to append the entry name to the project's .gitignore idempotently. vcs: tracked leaves .gitignore untouched.

---

Init (init:)

init: has no host-side marker (the declarative marker was removed — "Model A"). The run command is sourced inline in the attach shell before the harness starts, on every attach, so it must self-gate cheaply (e.g. bd list >/dev/null 2>&1 || bd init). There are no per-recipe marker files and no transient podman run --rm. See recipe authoring — One-time init.

---

Misc fixed mounts

Host path	Container path	Mode	Notes
catalog/base/egress-firewall.sh	/usr/local/sbin/egress-firewall	ro	Applied by _apply_firewall after the container starts; not configurable per recipe

---

See also

• Recipe authoring — persist: and init: schema reference (field names, valid scope/location combos, vcs:, worked examples)
• harnessed-design.md — the design rationale behind the mount model (profile-ahead-of-time, path mirroring, no-host-defaults invariant, secrets threat model)
• research/home-folder-claude-requirements.md — why the specific .claude/ subpaths are forwarded from the host (and which are intentionally excluded)
• Stacks — state.session_state in stack.yaml is a different, stack-level mechanism (where Claude session history volumes land: host vs volume) and is currently unimplemented in the launcher. Do not conflate it with recipe persist: entries, which are fully implemented and described on this page.
• Secrets (varlock + 1Password) — how secrets reach the container at launch time without being baked into the image or the profile