File tables and document viewing
ls displays names, types, sizes and timestamps in a colored table. A pipeline filters files, and show displays a document. The prompt also shows the directory and Git status.
Everyday commands, pipelines that work with fields, and Auto scripts you can edit and run in your terminal. The interactive session and CLI share the same execution engine.
Keep directory, variables and aliases between commands. Completion, history and inline suggestions reduce typing.
Query files and JSON, filter, sort and select fields. Use --json to pass results to another program.
Extend a command into functions, loops and conditions. Write multiline AutoScript in your terminal.
Two original captures show the prompt, readable tables and a continuous record of commands.
ls displays names, types, sizes and timestamps in a colored table. A pipeline filters files, and show displays a document. The prompt also shows the directory and Git status.
The Shell prompt keeps a sequence of directory and disk-usage commands. where and select work with fields, then to_json exports the result: terminal work and structured data in one session.
Switch between commands, script editing and AI chat. Each form has a separate image you can enlarge.
F1 locks input to shell commands. The right prompt reads Shell and the command prompt turns blue. Commands retain this mode; press F1 again to unlock.
F2 opens a numbered editor. Enter adds a line, F5 or Ctrl+Enter runs, and Esc cancels. Execution returns to the normal prompt. Three complete scripts follow below.
F3 enters an AI session. The right prompt reads AI and input uses ?. A model service is required; Esc returns to the shell.
The two captures above are original interface screenshots. These three images are redrawn from actual PTY interactions on 2026-09-30. F3 shows only its input area, excluding conversation history.
Each command returns to the prompt. Change directories, keep variables, load scripts and manage background jobs in the same session.
F3 requires model service configuration. These demonstrations made no external model calls. F4 is retired.
ls returns file records; from_json parses JSON. Filter and sort by fields without splitting text columns.
ls | .type == "dir" | select .name | first 5
cat users.json | from_json | .age > 30 | select .name .ageSave users.json in the current directory to run the query below. Download users.json ↗
F2 opens the multiline editor. Enter inserts a newline; F5 or Ctrl+Enter runs the script. Each run returns to the normal prompt.
Define a classification function and call it for each score in an array.
fn bucket(n) {
if n >= 80 { return "high" }
return "low"
}
var values = [95, 82, 61]
for v in values {
print(v.str() + ":" + bucket(v))
}95:high
82:high
61:lowF5 has a reliable terminal key sequence. Ctrl+Enter depends on your terminal reporting its modifiers. All three images record actual F2 editor runs submitted with F5, redrawn from PTY output.
Run session commands with ash -c, or save multiline code as a .ash file. CLI startup options control JSON output and execution policy.
--dry-run describes intended operations; --read-only rejects writes.
--sandbox confines built-in file operations and redirection. --allow and --no-exec control command execution.
--audit writes JSONL audit records; --no-network blocks the corresponding network commands.
ash --dry-run -c 'touch output.txt'
ash --read-only -c 'touch output.txt'
ash --audit ash-audit.jsonl -c 'pwd'Path policy applies to file access managed by ash. External programs need OS isolation for their own file access; combine policy with command restrictions.
Add the built ash executable to PATH, then choose your entry point.
Start ash and type commands at its prompt.
ashFor command calls from scripts and other programs.
ash --json -c 'ls | sort .size desc | select .name .size | first 5'Save the script and users.json in the same directory.
ash user-report.ashYou need the Rust toolchain and sibling auto-lang, auto-ai and auto-shell checkouts. Follow the repository instructions for dependencies; build the CLI from auto-shell/ash:
cargo build --release -p ash
cargo run --release -p ashThis page uses source builds and observed runs from 2026-09-29/30. The website release, AutoLang v0.5, and the executable version, ash v0.1.0, are identified separately.