feat: rework multi-agent plan + complete CLI surface (19/19 commands) #138

Merged

clawdie merged 3 commits from feat/multi-agent-plan-rework-cli-surface into main

2026-06-21 18:47:04 +02:00

clawdie commented

2026-06-21 18:39:51 +02:00

Owner

Summary

Reworks MULTI-AGENT-HOST-PLAN.md to reflect 0.11.0 reality and completes the CLI surface.

Plan doc rework

Context updated to released 0.11.0 (was 0.10.0 staged)
Command inventory: 19 (was 17), CLI surface: 19/19 (was 10/17)
Phase 2a marked COMPLETE (register-agent/list-agents merged)
Added Closed-gaps section documenting tenant/vault chain, issues #88/#91/#92
Applied positive language convention
Narrowed gap analysis to remaining open items

CLI surface completion (Phase 2b)

All 19 socket commands now have CLI wrappers:

claim-task --task-id UUID --agent-id UUID
transition-task --task-id UUID --status STATUS
set-cost-mode MODE

Remote agents can work entirely through the colibri binary instead of raw Python socket calls.

Phase 1a: pick_agent unit tests (3 new)

test_pick_agent_tie_breaking — equal score → later-in-slice wins (deterministic)
test_pick_agent_multiple_required_capabilities — agent with both caps beats one
test_pick_agent_active_status_eligible — active status eligible same as idle

Phase 2c: CLI parse tests (6 new)

Positive + negative tests for each new command.

Gate

cargo fmt --check — clean
cargo clippy --workspace --all-targets -- -D warnings — clean
cargo test --workspace — 245 passed, 0 failed, 1 ignored
cargo build --workspace --release — clean

(Sam & Claude)

## Summary Reworks `MULTI-AGENT-HOST-PLAN.md` to reflect 0.11.0 reality and completes the CLI surface. ### Plan doc rework - Context updated to released 0.11.0 (was 0.10.0 staged) - Command inventory: 19 (was 17), CLI surface: 19/19 (was 10/17) - Phase 2a marked COMPLETE (register-agent/list-agents merged) - Added Closed-gaps section documenting tenant/vault chain, issues #88/#91/#92 - Applied positive language convention - Narrowed gap analysis to remaining open items ### CLI surface completion (Phase 2b) All 19 socket commands now have CLI wrappers: - `claim-task --task-id UUID --agent-id UUID` - `transition-task --task-id UUID --status STATUS` - `set-cost-mode MODE` Remote agents can work entirely through the `colibri` binary instead of raw Python socket calls. ### Phase 1a: pick_agent unit tests (3 new) - `test_pick_agent_tie_breaking` — equal score → later-in-slice wins (deterministic) - `test_pick_agent_multiple_required_capabilities` — agent with both caps beats one - `test_pick_agent_active_status_eligible` — active status eligible same as idle ### Phase 2c: CLI parse tests (6 new) Positive + negative tests for each new command. ## Gate - `cargo fmt --check` — clean - `cargo clippy --workspace --all-targets -- -D warnings` — clean - `cargo test --workspace` — 245 passed, 0 failed, 1 ignored - `cargo build --workspace --release` — clean (Sam & Claude)

clawdie added 2 commits 2026-06-21 18:39:53 +02:00

docs: rework MULTI-AGENT-HOST-PLAN for 0.11.0 — narrowed gaps, Phase 2a done 4792ca84e4

Update context to released 0.11.0 (was 0.10.0 staged). Command inventory
now 19 (was 17), CLI surface 19/19 (was 10/17). Mark Phase 2a COMPLETE
(register-agent/list-agents merged). Add Closed-gaps section documenting
tenant/vault chain, issues #88/#91/#92. Apply positive language convention.
Narrow gap analysis to remaining open items.

(Sam & Claude)

feat(client): expose claim-task/transition-task/set-cost-mode in CLI

CI / markdown (pull_request) Has been cancelled

Details

CI / port (pull_request) Has been cancelled

Details

CI / agent-jail-pkgs (pull_request) Has been cancelled

Details

CI / rust (pull_request) Has been cancelled

Details

79637bab81

Completes the CLI surface: all 19 socket commands now have CLI wrappers.
Remote agents can work entirely through the colibri binary instead of raw
Python socket calls.

CLI additions:
- claim-task --task-id UUID --agent-id UUID
- transition-task --task-id UUID --status STATUS
- set-cost-mode MODE

DaemonClient gains claim_task(), transition_task(), set_cost_mode().

Also adds 3 pick_agent unit tests (Phase 1a):
- tie-breaking: equal score → later-in-slice wins (deterministic)
- multi-required-capabilities: agent with both caps beats one
- active-status-eligible: 'active' status eligible same as 'idle'

Plus 6 CLI parse tests (positive + negative) for the new commands.

Gate: fmt clean, clippy clean, 245 tests pass (Sam & Claude)