UAT: A2A workflow never exposes input-required task state #6992

New issue

Open

opened 2026-04-10 06:12:04 +00:00 by HAL9000 · 0 comments

HAL9000 commented

2026-04-10 06:12:04 +00:00

Owner

What was tested

Reviewed the Human-in-the-Loop collaboration and A2A task lifecycle sections (docs/specification.md lines 23506-23572) on commit 51aab18411.

Expected behavior

Tasks should transition into an input-required state whenever the server-hosted agent needs client-local resources or human approval.
The A2A implementation should surface that state so clients can resume the task after providing input, matching the spec's task state diagram.

Actual behavior

The codebase never references input-required or input_required. There is no task state or enum value representing this mode, so clients cannot be notified that input is required.
_cleveragents/fs/* and _cleveragents/terminal/* extensions have no support code to drive a task into an input-needed state; the workflow cannot match the spec.

Steps to reproduce

From the repo root, run:

python3 - <<'PY'
import pathlib
root = pathlib.Path('src/cleveragents')
hits = [str(p.relative_to(root)) for p in root.rglob('*.py')
        if 'input-required' in p.read_text(encoding='utf-8')
           or 'input_required' in p.read_text(encoding='utf-8')
           or 'INPUT_REQUIRED' in p.read_text(encoding='utf-8')]
print('matches:', hits)
PY

The script prints an empty list, confirming no task state implements the spec-required input-required mode.
Inspect src/cleveragents/a2a (transport, events, models) and plan lifecycle code — none define an input-required state.

Code location

src/cleveragents/a2a/ module (tasks, transport, events)
src/cleveragents/application/services/plan_lifecycle_service.py
src/cleveragents/agents/ (session/task state handling)

## What was tested - Reviewed the Human-in-the-Loop collaboration and A2A task lifecycle sections (docs/specification.md lines 23506-23572) on commit 51aab184112728471a44d5a91c334663cf8cd016. ## Expected behavior - Tasks should transition into an `input-required` state whenever the server-hosted agent needs client-local resources or human approval. - The A2A implementation should surface that state so clients can resume the task after providing input, matching the spec's task state diagram. ## Actual behavior - The codebase never references `input-required` or `input_required`. There is no task state or enum value representing this mode, so clients cannot be notified that input is required. - `_cleveragents/fs/*` and `_cleveragents/terminal/*` extensions have no support code to drive a task into an input-needed state; the workflow cannot match the spec. ## Steps to reproduce 1. From the repo root, run: ```bash python3 - <<'PY' import pathlib root = pathlib.Path('src/cleveragents') hits = [str(p.relative_to(root)) for p in root.rglob('*.py') if 'input-required' in p.read_text(encoding='utf-8') or 'input_required' in p.read_text(encoding='utf-8') or 'INPUT_REQUIRED' in p.read_text(encoding='utf-8')] print('matches:', hits) PY ``` 2. The script prints an empty list, confirming no task state implements the spec-required `input-required` mode. 3. Inspect `src/cleveragents/a2a` (transport, events, models) and plan lifecycle code — none define an input-required state. ## Code location - `src/cleveragents/a2a/` module (tasks, transport, events) - `src/cleveragents/application/services/plan_lifecycle_service.py` - `src/cleveragents/agents/` (session/task state handling)

Rows
Columns