To keep the backslashes in Windows file paths and stop `shlex.Split`
from interpreting them as escape characters, we need to quote the paths.
Signed-off-by: Babak K. Shandiz <babakks@github.com>
Expanded usage examples in the command help to clarify reading from stdin and selecting a base branch. Updated flag descriptions for --from-file and --base to provide clearer guidance to users.
Introduces a BaseBranch field to CreateOptions and updates the job creation flow to include the base branch in the payload if specified. The CapiClient and related methods are updated to accept and handle the baseBranch parameter, and tests are added to verify correct behavior when a base branch is provided.
Refactored file reading in agent-task create command to use cmdutil.ReadFile, allowing input from stdin. Updated tests to cover reading file content from stdin and improved test setup for file and stdin input cases.
The agent-task create command now accepts a task description from a file using the -F/--from-file flag, with mutual exclusivity enforced between inline and file input. Tests were updated to cover new input scenarios and error cases, and usage examples were added to the command help.
Replaces 'problem statement' with 'task description' in error messages for clarity. Refactors tests to use a BaseRepo function instead of direct repo objects, and adds a test for missing task description error.
Refactored the CreateJob method to decode the response body into the Job struct before checking for error status codes. Error messages are now extracted directly from the Job struct, removing redundant error parsing logic.
Removed unnecessary conditional check for wantStdout in Test_createRun. Now always asserts equality between wantStdout and actual stdout output, improving test clarity.
Normalizes job API error messages to include status code and text, ensures errors from job polling are surfaced to stderr without halting execution, and updates tests to verify error output and remove outdated fallback behavior.
Uses url.PathEscape for repo owner, repo name, and session ID when constructing agent session URLs to ensure proper encoding and prevent issues with special characters.
Simplifies the error message when a repository is missing by removing the unnecessary 'error:' prefix, making the output clearer and more user-friendly.
Changed the command usage string from 'create <problem statement>' to 'create "<task description>"' for improved clarity in the agent task creation command.
Changed argument parsing to accept a maximum of one argument and assign it directly to ProblemStatement. Added error handling for missing problem statement and removed unused strings import.
Introduces the 'create' subcommand for agent tasks, allowing users to create agent jobs via the Copilot API. Adds job API client methods, job model, and polling logic to retrieve associated pull requests. Includes tests for various job creation scenarios.
Introduces pager functionality to the agent-task list command output for improved readability. If the pager fails to start, an error message is printed to stderr. Aligns with implementations in other commands.
Added a test case to ensure that web mode in listRun uses the global URL even when the --repo flag is set. This improves coverage for scenarios where both web mode and repo are specified.
Added a comment explaining that the web GUI does not currently support filtering by repo, so the agents dashboard is opened without arguments. Notes future improvement if repo filtering becomes available.
Replaces plain output with a NoResultsError in listRun when no agent tasks are found. Updates related tests to expect the error instead of output, improving error handling consistency.
