UAT: `automation-profile show` JSON/YAML output and rich display missing 5 of 8 SafetyProfile sub-model fields #4861

New issue

Open

opened 2026-04-08 20:10:12 +00:00 by HAL9000 · 0 comments

HAL9000 commented

2026-04-08 20:10:12 +00:00

Owner

Summary

The CLI's automation-profile show command (and automation-profile list) omits 5 of the 8 SafetyProfile sub-model fields from both the structured JSON/YAML output and the rich terminal display. Only require_sandbox, require_checkpoints, and allow_unsafe_tools are shown; the remaining 5 fields are silently dropped.

What Was Tested

Code-level analysis of src/cleveragents/cli/commands/automation_profile.py against docs/reference/automation_profiles.md and ADR-041.

Expected Behavior (from spec)

The SafetyProfile sub-model has 8 fields per ADR-041 and docs/reference/automation_profiles.md:

Field	Type	Default
`require_sandbox`	bool	`true`
`require_checkpoints`	bool	`true`
`allow_unsafe_tools`	bool	`false`
`require_human_approval`	bool	`false`
`allowed_skill_categories`	list[str]	`[]`
`max_cost_per_plan`	float \| null	`null`
`max_total_cost`	float \| null	`null`
`max_retries_per_step`	int	`3`

All 8 fields should appear in the safety sub-object of the JSON/YAML output and in the rich display.

Actual Behavior (from code)

`_profile_spec_dict()` — JSON/YAML output

# src/cleveragents/cli/commands/automation_profile.py
"execution_controls": {
    "install_dependency": profile.install_dependency,
    "require_sandbox": profile.safety.require_sandbox,
    "require_checkpoints": profile.safety.require_checkpoints,
    "allow_unsafe_tools": profile.safety.allow_unsafe_tools,
    # MISSING: require_human_approval, allowed_skill_categories,
    #          max_cost_per_plan, max_total_cost, max_retries_per_step
},

The 5 missing safety fields are: require_human_approval, allowed_skill_categories, max_cost_per_plan, max_total_cost, max_retries_per_step.

`_print_profile()` — Rich display

f"  require_sandbox: {profile.safety.require_sandbox}\n"
f"  require_checkpoints: {profile.safety.require_checkpoints}\n"
f"  allow_unsafe_tools: {profile.safety.allow_unsafe_tools}"
# MISSING: require_human_approval, allowed_skill_categories,
#          max_cost_per_plan, max_total_cost, max_retries_per_step

Same 5 fields are absent from the rich terminal panel.

Impact

Users cannot inspect the full safety configuration of a profile via CLI
Custom profiles that set max_cost_per_plan, max_total_cost, require_human_approval, or allowed_skill_categories will appear to have no such constraints when shown
The max_retries_per_step value (default 3) is never visible in CLI output
JSON/YAML output is incomplete for programmatic consumers

Code Location

src/cleveragents/cli/commands/automation_profile.py
- _profile_spec_dict() function — execution_controls dict
- _print_profile() function — rich panel details string

Steps to Reproduce

Run agents automation-profile show manual --format json
Observe the output does not include require_human_approval, allowed_skill_categories, max_cost_per_plan, max_total_cost, or max_retries_per_step in the safety section
Run agents automation-profile show manual
Observe the rich panel does not display these 5 fields

Definition of Done

_profile_spec_dict() includes all 8 SafetyProfile fields in the output (either nested under safety: key or in execution_controls)
_print_profile() displays all 8 SafetyProfile fields in the rich panel
Tests verify the complete safety sub-model appears in JSON, YAML, and rich output

Automated by CleverAgents Bot
Supervisor: UAT Testing | Agent: uat-tester

## Summary The CLI's `automation-profile show` command (and `automation-profile list`) omits 5 of the 8 `SafetyProfile` sub-model fields from both the structured JSON/YAML output and the rich terminal display. Only `require_sandbox`, `require_checkpoints`, and `allow_unsafe_tools` are shown; the remaining 5 fields are silently dropped. ## What Was Tested Code-level analysis of `src/cleveragents/cli/commands/automation_profile.py` against `docs/reference/automation_profiles.md` and ADR-041. ## Expected Behavior (from spec) The `SafetyProfile` sub-model has **8 fields** per ADR-041 and `docs/reference/automation_profiles.md`: | Field | Type | Default | |-------|------|---------| | `require_sandbox` | bool | `true` | | `require_checkpoints` | bool | `true` | | `allow_unsafe_tools` | bool | `false` | | `require_human_approval` | bool | `false` | | `allowed_skill_categories` | list[str] | `[]` | | `max_cost_per_plan` | float \| null | `null` | | `max_total_cost` | float \| null | `null` | | `max_retries_per_step` | int | `3` | All 8 fields should appear in the `safety` sub-object of the JSON/YAML output and in the rich display. ## Actual Behavior (from code) ### `_profile_spec_dict()` — JSON/YAML output ```python # src/cleveragents/cli/commands/automation_profile.py "execution_controls": { "install_dependency": profile.install_dependency, "require_sandbox": profile.safety.require_sandbox, "require_checkpoints": profile.safety.require_checkpoints, "allow_unsafe_tools": profile.safety.allow_unsafe_tools, # MISSING: require_human_approval, allowed_skill_categories, # max_cost_per_plan, max_total_cost, max_retries_per_step }, ``` The 5 missing safety fields are: `require_human_approval`, `allowed_skill_categories`, `max_cost_per_plan`, `max_total_cost`, `max_retries_per_step`. ### `_print_profile()` — Rich display ```python f" require_sandbox: {profile.safety.require_sandbox}\n" f" require_checkpoints: {profile.safety.require_checkpoints}\n" f" allow_unsafe_tools: {profile.safety.allow_unsafe_tools}" # MISSING: require_human_approval, allowed_skill_categories, # max_cost_per_plan, max_total_cost, max_retries_per_step ``` Same 5 fields are absent from the rich terminal panel. ## Impact - Users cannot inspect the full safety configuration of a profile via CLI - Custom profiles that set `max_cost_per_plan`, `max_total_cost`, `require_human_approval`, or `allowed_skill_categories` will appear to have no such constraints when shown - The `max_retries_per_step` value (default 3) is never visible in CLI output - JSON/YAML output is incomplete for programmatic consumers ## Code Location - `src/cleveragents/cli/commands/automation_profile.py` - `_profile_spec_dict()` function — `execution_controls` dict - `_print_profile()` function — rich panel details string ## Steps to Reproduce 1. Run `agents automation-profile show manual --format json` 2. Observe the output does not include `require_human_approval`, `allowed_skill_categories`, `max_cost_per_plan`, `max_total_cost`, or `max_retries_per_step` in the safety section 3. Run `agents automation-profile show manual` 4. Observe the rich panel does not display these 5 fields ## Definition of Done - [ ] `_profile_spec_dict()` includes all 8 `SafetyProfile` fields in the output (either nested under `safety:` key or in `execution_controls`) - [ ] `_print_profile()` displays all 8 `SafetyProfile` fields in the rich panel - [ ] Tests verify the complete safety sub-model appears in JSON, YAML, and rich output --- **Automated by CleverAgents Bot** Supervisor: UAT Testing | Agent: uat-tester

HAL9000 referenced this issue

2026-04-08 20:11:32 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue

2026-04-08 20:16:04 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 added this to the v3.5.0 milestone

2026-04-08 20:18:55 +00:00

HAL9000 referenced this issue

2026-04-08 20:19:20 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue