yana-rt 1.4.0

Yana AI Runtime — safety CLI for AI agents: scan, graph, vault, hunt, ci, map, fix, doctor
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

//! Chat adapter for canonical command validation and execution.

use std::path::Path;

pub use crate::capability::{CommandOutcome as ExecOutcome, ValidatedCommand as Validated};

/// Pure, synchronous: parses the command into argv and asks
/// `crate::guard::check_command()` — the single source of judgment for
/// "is this destructive" (identical logic to
/// `core/hooks/guard-destructive.sh`) — never a second pattern list of
/// its own.
pub fn validate(command: &str) -> Result<Validated, String> {
    crate::capability::validate_command(command).map_err(Into::into)
}

/// Actually runs the command. Only ever called after (a) `validate()`
/// found `guard_verdict == None`, and (b) the human approved via the TUI
/// — this function does not re-check either condition itself. Direct
/// argv exec, no `sh -c` string-building (per `shell-sanitize-law.md`,
/// same style as `plugin.rs`/`guard/lock.rs`). When `use_sandbox` is
/// true, routes through `core/scripts/sandbox-exec.sh` for real
/// Docker/nsjail/ulimit isolation instead of running the argv directly.
pub fn execute(
    repo_root: &Path,
    argv: &[String],
    use_sandbox: bool,
) -> Result<ExecOutcome, String> {
    crate::capability::execute_command(repo_root, argv, use_sandbox).map_err(Into::into)
}