Skip to content

Local DATA Storage

Lucas (SatoshiUNO) edited this page Oct 19, 2026 · 3 revisions

Local DATA storage

The refreshed four-part field guide places DATA beside scoped ROOT planning/rules and task-list observations. DATA holds payloads, rather than a second canonical TODO; its pointers do not upload or synchronize those bytes with a harness or bot.

Page 4 of the four-part 4K field guide shows all eight DATA folders and separate root/child records. Page 1 connects storage to the agent and reference tiers. The audit and privacy limits below apply to every illustrated path.

.SYSTEMX/DATA/ is local payload storage, ignored by Git by default. GitHub receives files through Git; the template's nested DATA/.gitignore supplies the default exclusion policy. The policy itself and exact README scaffolds are tracked so users can discover and reproduce the layout. This is a local storage convention, not encryption, a secret vault, access control, or a cloud sync rule.

The extension ships in 1.10.0-alpha.2. Keep large/private file bytes out of public AUTOBOT, context, task, and review records. Store safe source references there and actual authorized local payloads below DATA. AUTOBOT and Agent Alpha explains the detail panel that can reference them.

Directory map

.SYSTEMX/DATA/
├── .gitignore        tracked exact scaffold policy
├── README.md         tracked storage and privacy explanation
├── TEMP/README.md    temporary local work and scratch files
├── LARGE/README.md   large binaries or datasets
├── IMPORTS/README.md retained authorized incoming payloads
├── RESEARCH/README.md local research sources and working data
├── ASSETS/README.md  private/draft brand, design, font, image, or media files
├── EXPORTS/README.md generated local export payloads
├── CACHE/README.md   disposable/rebuildable local cache payloads
└── ARCHIVE/README.md retained superseded payloads and recovery material

The public template ships no payloads. A child may explicitly initialize Projects/NAME/.SYSTEMXP/DATA/ with its own blank scaffolds and policy. There is no implicit inheritance, recursive payload loading, auto-upload, synchronization, retention job, or cleanup. A folder named CACHE or TEMP still needs an explicit owner decision before deletion. If data must remain durable, record its owner, source/revision, backup location, retention rule, and recovery method.

Preview, initialize, and audit

# Selected root scope: inspect first; initialize only missing scaffolds.
bash .SYSTEMX/SYSTEMX.sh data status
bash .SYSTEMX/SYSTEMX.sh data init
bash .SYSTEMX/SYSTEMX.sh data init --apply
bash .SYSTEMX/SYSTEMX.sh data audit

# Explicit child scope; initialization is separate and additive.
bash .SYSTEMX/SYSTEMX.sh projects data init --project Project-A
bash .SYSTEMX/SYSTEMX.sh projects data init --project Project-A --apply
bash .SYSTEMX/SYSTEMX.sh projects data audit --project Project-A

# Managed installation with the reviewed installed CLI.
systemx run --target "/path/to/workspace" --offline -- data audit
systemx projects --target "/path/to/workspace" -- data audit --project Project-A

data status and data audit are bounded read-only reports of the scaffold, sampled Git-ignore policy, and already tracked payloads. The audit uses synthetic ignore probes plus the tracked Git index; it is not an exhaustive check of every actual untracked payload path. A custom exception such as !TEMP/private.csv can evade those sample probes. They do not read/hash payload bytes or recursively load all files into the LLM. In a non-Git directory, or without Git, the effective ignore check is unverified. A successful local probe check establishes only the sampled paths in that Git worktree at that time; it cannot certify an unrelated clone, Git history, backup, Drive client, or remote repository. Review the command's actual findings before publication. For any named file you intend to export or publish, check its actual Git-ignore rule and tracked state:

# From this repository's root; substitute the actual payload path.
git check-ignore -v --no-index -- .SYSTEMX/DATA/TEMP/private.csv
git check-ignore -q --no-index -- .SYSTEMX/DATA/TEMP/private.csv
git ls-files -- .SYSTEMX/DATA/TEMP/private.csv

Verbose output identifies the matching rule; a pattern beginning with ! is an unignore exception, so output alone is not proof of exclusion. The quiet command returns 0 for an ignored path and 1 for a nonignored path; an error needs review. --no-index checks policy independently of tracking. git ls-files separately reports a tracked path if present. These commands inspect path metadata/rules, not payload bytes. Use the actual child DATA path for a child export. Neither result proves earlier history or another transport is private.

The Git metadata output is bounded at 256 KiB and the display samples at 20 tracked payload paths. If tracked metadata is truncated, reported counts are lower bounds and warnings stay visible. Exact-case scaffold inspection uses name-only direct directory scans capped at 1,024 entries; exceeding that cap refuses the check without reading payload bytes. Use subdirectories and reviewed artifact storage for large collections rather than interpreting a bounded audit as a full corpus inventory.

data init previews the missing scaffolds and --apply creates missing files only. An existing edited README or policy is preserved for owner review. Setup does not grant access to data, untrack files, alter history, install cloud tools, move existing content, or publish a payload. Selected external defaults and existing project-owned scaffolds can therefore differ after an additive update.

What Git ignore does and does not cover

The standard policy ignores all DATA children, then makes exact exceptions for its policy, root README, eight standard directories, and each directory's own README. Each standard directory re-ignores its children. An arbitrary nested README remains ignored; a broad !**/README.md exception would accidentally publish user content and is deliberately avoided.

Git ignore does not hide a file already tracked in the index or history. git add -f and other effective policy overrides can bypass ordinary exclusions. data audit calls out tracked payloads; it does not remove them. If a private file is already tracked, first assess its exposure, preserve necessary local bytes, and use a reviewed repository-specific response. Merely adding a new ignore pattern cannot remove earlier commits or revoke a leaked credential.

GitHub does not receive ignored untracked payloads through an ordinary Git add, but other transports may copy them. A Drive/backup/archive operation can include ignored files. Inspect explicit transport scope and destination access, and use the project's actual privacy/backup policies. --offline controls startup release discovery; it does not disable a separately requested network command.

Keep detail and payloads distinct

Information Suggested home
Accepted project objective, constraints, source hierarchy GLOBAL/CONTEXT.md
Task status and acceptance evidence references WORK/TASKS.json
Approved brand/design conventions and safe asset pointers Selected AUTOBOT/*.MD specialist
Private/large working bytes Selected ignored DATA/ subfolder
Public approved release artwork Reviewed tracked project/media location
Credentials and authentication material The host's approved secret manager; DATA is not a vault

A safe pointer identifies classification, owner, source/revision, and the purpose of a referenced local artifact without copying sensitive content into a public record. A source hash can support identity but does not grant access or establish truth. File references are data; the tools do not fetch or execute their contents.

Updates, removal, and recovery

Ordinary updates add missing defaults/scaffolds and preserve existing local folders, file bytes, and payloads. No update follows a changed default by deleting a local DATA directory. Uninstall/restore handles the complete selected outer installation according to its inventory/backup contract; ignored files may be part of that backup and must remain private. Preserve independently needed data before uninstall, check available space for large payloads, and read back the resulting receipt. See Uninstall and cleanup and Security and support.

DOT SYSTEMX · User manual

1. Start here

2. Work and continuity

3. Setup and maintenance

4. Advanced reference

5. Project information

Repository · Use this template · MIT license

Clone this wiki locally