Add PlanApplyService with diff(), artifacts(), persist_apply_summary(),
handle_merge_failure(), and guard_empty_changeset() methods.
- plan diff: renders changeset in rich/plain/json/yaml formats
- plan artifacts: shows plan metadata, changeset summary, sandbox refs
- persist_apply_summary: stores files_changed/validations_run in plan metadata
- handle_merge_failure: transitions plan to error state with conflict details
- guard_empty_changeset: blocks apply on empty changeset (--allow-empty override)
CLI: plan diff and plan artifacts commands with --format flag.
Tests: 20 Behave scenarios, 8 Robot integration tests, 4 ASV benchmarks.
Docs: docs/reference/plan_apply.md with CLI reference.
Implements D0b.apply from the implementation plan.
Mark as complete the post-commit quality verification and Forgejo PR
items for four commit groups whose code was already merged to master
via develop-brent-2 PR #132:
- C1.tool.domain.tests (feature/m3-tool-domain-robot): Forgejo PR
- C0.skill.domain.tests (feature/m3-skill-domain-robot): Forgejo PR
- 10C.edge (feature/m6-validation-edge): coverage, nox, Forgejo PR
- 10C.performance (feature/m6-perf-scale): coverage, nox, Forgejo PR
All quality checks were validated on develop-brent-2: lint, typecheck,
4437 behave scenarios, 515 robot tests, 97% coverage. Confirmed again
on develop-brent-3 (this branch).
Typer 0.24.0 CliRunner does not support the mix_stderr keyword
argument. This caused ASV benchmark discovery to fail with a
TypeError during the CI benchmark-regression job.
- Remove AutomationLevel imports from cli_robot_flow_bench.py and
persistence_robot_bench.py (enum was removed by master's automation
refactor); replace with AutomationProfileRef where needed.
- Use typer.echo() instead of console.print() for machine-readable
output (JSON/YAML/plain) in config.py and session.py to prevent
Rich from injecting ANSI escape codes that corrupt json.loads().
- Set NO_COLOR=1 in noxfile unit_tests, integration_tests, and
coverage_report sessions as a belt-and-suspenders safeguard for
all CLI commands that route format_output through Rich.
Add python -m compileall -q src/ step to the integration_tests nox
session before launching pabot. On CI runners with high core counts
(e.g. 32 processes) the first Robot test to spawn a Python subprocess
could fail because 30+ processes simultaneously cold-compile the
entire source tree from scratch. Pre-compiling eliminates the
thundering-herd race and lets all workers read cached .pyc files.
Master's refactor(automation) commit removed the AutomationLevel enum
and automation_level field from Plan. Update all behave steps, robot
helpers, and feature scenarios that referenced the removed type:
- cli_lifecycle_robot_alignment_steps.py: drop import and field assignment
- persistence_robot_alignment_steps.py: drop import and field assignment
- subplan_model_steps.py: drop import, parameter, and field assignment
- subplan_model.feature: replace automation level scenario with profile check
- helper_cli_lifecycle.py: drop import and field assignment
- helper_persistence_lifecycle.py: drop import, field assignment, and assertion
Renamed 'listing actors with namespace' step in actor_loading_steps.py
to 'listing loaded actors with namespace' to avoid collision with
actor_registry_persistence_steps.py from master. Updated
actor_loading.feature to use renamed step.
Add CLI command group `agents automation-profile` with add, remove,
list, and show subcommands for managing automation profiles that
control plan execution autonomy.
Changes:
- New `automation_profile.py` CLI module with YAML config input,
schema_version guard, namespaced name validation, --update support,
and all output formats (json/yaml/plain/table/rich)
- Register automation-profile in main CLI app
- Add deprecation warnings for --automation-level on plan use and
set-automation-level commands
- Update CLI reference docs with command examples, built-in profiles
list, and deprecation notes
- 26 Behave scenarios covering all commands and error paths
- 9 Robot Framework integration smoke tests
- ASV benchmarks for CLI parsing performance
- Mark A6.cli items complete in implementation_plan.md
Add tree, inspect, link-child, and unlink-child subcommands to the
agents resource CLI group. Extend ResourceRegistryService with
link_child, unlink_child, get_children, and get_resource_tree methods.
- tree: display resource hierarchy with --depth, --type, and format options
- inspect: show resource details with --tree and --file options
- link-child/unlink-child: manage DAG edges with cycle detection
- 24 Behave scenarios, 4 Robot integration tests, ASV benchmarks
- CLI reference documentation in docs/reference/resource_cli.md
CoreRasurae
freemo
brent.edwards