Files
cleveragents-core/features/tui_shell_exec_coverage.feature
HAL9000 8e8f2d5f89 fix(tui): enforce shell safety gating
- honour ShellSafetyService verdicts before executing shell commands
- tighten TUI confirmation defaults and align warning messaging with spec
- split shell-safety Behave steps and add Robot coverage

ISSUES CLOSED: #6361
2026-06-17 00:21:34 -04:00

48 lines
2.2 KiB
Gherkin

Feature: TUI Shell Exec Coverage
Scenarios that exercise previously uncovered code paths
in the tui/input/shell_exec module (lines 48-49, 52-56, 61, 77-82).
Background:
Given the shell_exec module is imported
Scenario: Empty command returns error result
When I run a shell command with an empty string
Then the shell result exit code should be 2
And the shell result stderr should be "empty command"
And the shell result stdout should be empty
Scenario: Whitespace-only command returns error result
When I run a shell command with only whitespace " "
Then the shell result exit code should be 2
And the shell result stderr should be "empty command"
Scenario: Shell mode disabled via env var set to 1
Given the environment variable CLEVERAGENTS_DISABLE_SHELL_MODE is set to "1"
When I run a shell command "echo hello"
Then the shell result exit code should be 2
And the shell result stderr should be "shell mode is disabled"
Scenario: Shell mode disabled via env var set to true
Given the environment variable CLEVERAGENTS_DISABLE_SHELL_MODE is set to "true"
When I run a shell command "echo hello"
Then the shell result exit code should be 2
And the shell result stderr should be "shell mode is disabled"
Scenario: Dangerous command confirmed via callback proceeds to execution
Given a confirm_dangerous callback that returns True
When I run a dangerous command "rm -rf /" with the callback
Then the dangerous command should have been confirmed
And the shell result should reflect the executed command
Scenario: Dangerous command confirmed but callback returns False is still blocked
Given a confirm_dangerous callback that returns False
When I run a dangerous command "rm -rf /" with the callback
Then the shell result exit code should be 1
And the shell result stderr should be "blocked by shell safety policy"
Scenario: Command that exceeds timeout returns timeout result
Given subprocess run is mocked to raise TimeoutExpired
When I run a shell command "sleep 999" with a 1 second timeout
Then the shell result exit code should be 124
And the shell result stderr should contain "command timed out after"