@openlife/cli 1.7.4 → 1.7.6

package/docs/audit/TEST-COVERAGE-GAPS.md

@@ -1,334 +0,0 @@
-# Test Coverage Gaps — openlife-cli
-
-**Source of truth:** `docs/audit/CLI-MATRIX.md` (125 leaf commands, 156 raw `.command()` registrations)
-**Tests scanned:** 56 `src/test_*.ts` files
-**Generated:** 2026-05-11
-
-This report counts only **leaf commands** (those a user actually invokes). Group-only `.command()` declarations (`auth`, `models`, etc.) are not counted as commands to cover.
-
----
-
-## Summary
-
-| Coverage level | Count | % |
-|----------------|-------|---|
-| Direct (subprocess spawn against `bin/openlife.js` or `dist/index.js`) | 36 | 28.8% |
-| Indirect (test covers underlying class/subsystem only) | 28 | 22.4% |
-| None | 61 | 48.8% |
-| **Total leaf commands** | **125** | **100%** |
-
-Hot/cold group summary:
-
-- **Hottest groups (best E2E coverage):** `plugin` (5/5 direct), `context` (3/3 direct), `pilot` (4/4 direct), `eval` (2/2 direct), `learn` (2/2 direct), `service` (6/6 direct).
-- **Coldest groups (zero coverage):** `memory` (5/5 none), `governance` (4/5 none after policy show), `benchmark` (3/3 none in CLI; subsystem only), `runtime` (2/2 indirect), `media` (3/4 none, only `media doctor` covered).
-
----
-
-## Commands with NO COVERAGE
-
-(61 commands, grouped by command group)
-
-### Top-level standalone (5 of 11 with no coverage)
-- `install` (line 200) — no test invokes `openlife install`; `test_install_flow.ts` tests the underlying `InstallFlow` class only (indirect, listed below)
-- `phase1-check` (437) — no test spawns the command
-- `restart` (1171) — no test
-- `update` (1186) — no test
-- `up` (1204) — no test
-- `start` (1270) — no test invokes `start --daemon` end-to-end (Gateway is tested separately in `test_daemon_sigterm.ts` and `test_trigger_basic_auth.ts`, but those instantiate `Gateway` directly without going through the CLI)
-
-> **Top-level covered**: `ask` (direct via `test_ask_exit.ts` + `test_openlife_gatekeeper_routing.ts`), `agent <verb>` (none, see below), `dream` (none — see below), `status` (none), `doctor` (none)
-
-Adding the unmentioned:
-- `agent <verb>` (1248) — top-level alias command, no direct test
-- `dream <entry...>` (alias `/dream`) (1345) — no CLI spawn; underlying `DreamOrganizer` class is covered by `test_dream_organizer.ts` (indirect)
-- `status` (1150) — no test
-- `doctor` (1163) — no test
-
-### `auth` group
-- `auth gemini` (246) — no test
-- `auth openai` (alias `codex`) (259) — no test
-
-### `models` group
-- `models status` (304) — no test
-- `models set <provider/model>` (323) — no test
-- `models fallbacks list` (339) — no test
-- `models fallbacks add <provider/model>` (348) — no test
-- `models fallbacks remove <provider/model>` (369) — no test
-- `models fallbacks clear` (384) — no test
-
-> Note: `Brain` fallback **logic** is tested by `test_brain_fallback_chain.ts` and `test_brain_error_diagnostics.ts`, but the `models` CLI surface that mutates `models.json` is untested.
-
-### `system` group (4 of 7 no coverage)
-- `system install` (467) — no test
-- `system setup` (505) — no test (alias of `system install`)
-- `system status` (535) — no test
-- `system doctor` (542) — no test
-- `system init-project` (552) — no test
-- `system doctor-world` (560) — no test
-- `system bootstrap-world` (567) — no test
-
-> Subsystem coverage exists: `test_install_flow.ts` tests `InstallFlow` (indirect).
-
-### `memory` group (5/5 none, all indirect-only)
-- `memory status` (573)
-- `memory provider-list` (577)
-- `memory provider-use <name>` (581)
-- `memory test <name>` (585)
-- `memory search <query...>` (589)
-
-### `swarm` group (1/2 — `swarm run` covered, `consensus-test` not)
-- `swarm consensus-test` (621) — no test
-
-### `governance` group (4/5 — only `policy show` covered)
-- `governance status` (627) — no test
-- `governance audit` (631) — no test
-- `governance risk-check <goal>` (635) — no test (underlying `GovernanceLayer.evaluate` is tested in `test_governance.ts`)
-- `governance consent <scope> <userId>` (639) — no test (underlying `GovernanceConsentStore` is tested in `test_governance_advanced.ts`)
-
-### `benchmark` group (3/3 none at CLI level; subsystem covered)
-- `benchmark run` (669)
-- `benchmark compare` (673)
-- `benchmark report` (677)
-
-> `BenchmarkEngine` class is exercised by `test_benchmark_engine.ts` (indirect).
-
-### `agents` group (1/3 — `list`/`show`/`create` covered)
-All three commands are covered at the CLI level (see Direct section).
-
-### `squads` group (1/2 — `create` direct, `list` not)
-- `squads list` (740) — no direct test (subsystem `SquadRegistry` is covered by `test_openlife_runtime_source_truth.ts`)
-
-### `skills` group (covered — `list` direct, `create` direct)
-
-### `route` group (1/1 covered)
-
-### `mcp` group (1/2 — `create` covered, `status` not)
-- `mcp status` (1137) — no test
-
-### `catalog` group (1/2 — `doctor` covered, `list` not)
-- `catalog list` (952) — no test
-
-### `teams` group (1/2 — `upsert` covered, `list` not)
-- `teams list` (958) — no direct test; subsystem covered by `test_agent_team_skill_network.ts`
-
-### `networks` group (1/2 — `upsert` covered, `list` not)
-- `networks list` (986) — no direct test; subsystem covered by `test_agent_team_skill_network.ts`
-
-### `media` group (1/4 — `doctor` covered)
-- `media route` (1013) — no test
-- `media route-status` (1027) — no test
-- `media read-image <imagePath>` (1083) — no test
-
-### `orchestration` group (2/2 none)
-- `orchestration contract` (1126) — no test
-- `orchestration snapshot` (1130) — no test
-
-### `smoke` group (2/2 none)
-- `smoke railway` (1324) — no test
-- `smoke telegram` (1331) — no test (note: `test_smoke*.ts` files do not exist)
-
-### `task` group (2/3 covered — `run` and `status`; `list` not)
-- `task list` (1434) — no test
-
-### `job` group (2/2 none at CLI; subsystem covered)
-- `job list` (1470)
-- `job events <jobId>` (1477)
-
-> `JobLifecycleStore` is covered by `test_job_lifecycle.ts` (indirect).
-
-### `runtime` group (2/2 none at CLI; subsystem covered)
-- `runtime probe` (1483)
-- `runtime list` (1487)
-
-> `RuntimeProbe` covered by `test_runtime_probe.ts`; `RuntimeRegistry` covered by `test_runtime_registry.ts` (both indirect).
-
-### `reversa` group (6/6 none at CLI; subsystem heavily covered)
-- `reversa status` (1493)
-- `reversa next` (1499)
-- `reversa mode` (1506)
-- `reversa run-phase <phase>` (1516)
-- `reversa run-all` (1522)
-- `reversa export-json` (1526)
-
-> `ReversaAgent` is heavily covered by `test_reversa_lite.ts`, `test_reversa_contracts_e2e.ts`, `test_reversa_export_and_strict.ts`, `test_reversa_full_execution.ts` (all indirect — they call `new ReversaAgent()` directly rather than spawning the CLI). `test_reversa_export_and_strict.ts` does spawn `aiobuilder generate-ui` which uses Reversa output, but no test spawns a `reversa <subcommand>` invocation.
-
-### `sources` group (2/4 covered — `scaffold` and `import-ref`)
-- `sources list` (1532) — no test
-- `sources guard-check <url>` (1536) — no test (underlying `SecurityDownloadGuard` covered by `test_security_download_guard.ts`)
-
-### `teammate` group (1/2 none at CLI; subsystem covered)
-- `teammate list` (1579) — no test (subsystem covered by `test_teammate_learning.ts`)
-- `teammate upsert <id> <owner> <status> <title>` (1586) — no test (subsystem covered)
-
-### `learning` group (2/2 none at CLI; subsystem covered)
-- `learning add <taskId> <summary>` (1593) — no test (subsystem `SkillLearningLoopStore` covered by `test_teammate_learning.ts`)
-- `learning list` (1597) — no test
-
-### `designmd` group (5/5 none at CLI; subsystem heavily covered)
-- `designmd status` (1691)
-- `designmd import` (1698)
-- `designmd list` (1711)
-- `designmd show <profileId>` (1715)
-- `designmd apply <profileId> <source> <title> <designPath>` (1720)
-
-> `DesignMdMode` / `DesignMdImporter` / `DesignMdRegistry` covered by `test_designmd_mode.ts`, `test_designmd_mode_workspace.ts`, `test_designmd_import_registry.ts` (all indirect).
-
-### `aiobuilder` group (2/6 covered — `validate-catalog`, `generate-ui`)
-- `aiobuilder mode` (1730) — no test
-- `aiobuilder create-agent <id>` (1741) — no test (underlying entity creation is tested by `test_create_entities.ts` via the `agents create` CLI command, but not the aiobuilder alias)
-- `aiobuilder create-squad <id>` (1749) — no test (same)
-- `aiobuilder chat <mensagem...>` (1845) — no test
-
----
-
-## Commands with INDIRECT COVERAGE ONLY
-
-28 commands. The test exists for the underlying class but does not exercise the CLI handler.
-
-| Command | Closest test | What it covers | What's missing |
-|---------|-------------|----------------|----------------|
-| `install` (top-level) | `test_install_flow.ts` | `InstallFlow.run()` against temp dir | CLI `--mode`, `--resume`, `--from-file`, `--headless`, `--models`, `--telegram-*` flag wiring |
-| `dream <entry...>` | `test_dream_organizer.ts` | `DreamOrganizer.organize()` classification | CLI invocation, output JSON shape, `/dream` alias |
-| `models status` | `test_brain_fallback_chain.ts`, `test_brain_error_diagnostics.ts` | Brain fallback logic | `models.json` reads, output formatting |
-| `models set <provider/model>` | `test_brain_fallback_chain.ts` (loosely) | Provider/model parsing happens in `ModelManager` (not directly tested at unit level) | CLI mutates `models.json`; no test |
-| `models fallbacks {list,add,remove,clear}` | (none direct) | `ModelManager` exists but no dedicated test | All four subcommands |
-| `memory status` | `test_memory_orchestrator.ts` | `MemoryOrchestrator.save/search` against local provider | CLI output, provider-registry status surface |
-| `memory provider-list` | `test_memory_orchestrator.ts` | Provider registry instantiation | CLI listing path |
-| `memory provider-use <name>` | (none) | — | Provider switching CLI |
-| `memory test <name>` | `test_memory_orchestrator.ts` | Local provider write/read | Connectivity test path |
-| `memory search <query...>` | `test_memory_orchestrator.ts`, `test_memory_promotion.ts`, `test_memory_retention.ts`, `test_conversation_memory.ts` | Search at API level | CLI dispatch + output |
-| `governance status` | `test_governance.ts`, `test_governance_advanced.ts` | `GovernanceLayer.evaluate` + `GovernanceConsentStore` | CLI status reporting |
-| `governance audit` | `test_governance_advanced.ts` | `SecurityEventStore` | CLI audit summary |
-| `governance risk-check <goal>` | `test_governance.ts`, `test_service_guardrails_delete.ts` | `evaluate()` blocks destructive/injection | CLI surface |
-| `governance consent <scope> <userId>` | `test_governance_advanced.ts` | `GovernanceConsentStore.approve` | CLI consent grant path |
-| `benchmark run` | `test_benchmark_engine.ts` | `BenchmarkEngine.record` + scoring | CLI runner |
-| `benchmark compare` | `test_benchmark_engine.ts` | Engine handles multiple labels | CLI compare report |
-| `benchmark report` | `test_benchmark_engine.ts` | Reporting at engine level | CLI report output |
-| `squads list` | `test_openlife_runtime_source_truth.ts` | `SquadRegistry.all()` source-truth check | CLI dispatch + count assertions (note: `agents list` IS covered directly; squads is the asymmetry) |
-| `teams list` | `test_agent_team_skill_network.ts` | `AgentTeamRegistry.upsert/resolve` | CLI list path |
-| `networks list` | `test_agent_team_skill_network.ts` | `SkillNetworkRegistry.upsert/resolve` | CLI list path |
-| `job list` / `job events` | `test_job_lifecycle.ts` | `JobLifecycleStore.enqueue/transition/events` | CLI dispatch |
-| `runtime probe` | `test_runtime_probe.ts` | `RuntimeProbe.probeAll()` covers 4 executors | CLI dispatch + JSON shape |
-| `runtime list` | `test_runtime_registry.ts` | `RuntimeRegistry.probe/list` | CLI dispatch |
-| `reversa {status,next,mode,run-phase,run-all,export-json}` | `test_reversa_lite.ts`, `test_reversa_full_execution.ts`, `test_reversa_contracts_e2e.ts`, `test_reversa_export_and_strict.ts` | Heavy `ReversaAgent` coverage: phases, modes, export | None of the 6 CLI handlers is exercised via subprocess |
-| `teammate list` / `teammate upsert` | `test_teammate_learning.ts` | `TeammateBoardStore` CRUD | CLI dispatch |
-| `learning add` / `learning list` | `test_teammate_learning.ts` | `SkillLearningLoopStore` CRUD | CLI dispatch |
-| `designmd {status,import,list,show,apply}` | `test_designmd_mode.ts`, `test_designmd_mode_workspace.ts`, `test_designmd_import_registry.ts` | Importer + Registry + Mode | None of the 5 CLI handlers is exercised |
-| `aiobuilder create-agent` / `aiobuilder create-squad` | `test_create_entities.ts` (via `agents create`/`squads create`) | Catalog file creation works | The aiobuilder-prefixed handlers specifically |
-| `start --daemon` | `test_daemon_sigterm.ts`, `test_trigger_basic_auth.ts` | Gateway listen + SIGTERM, basic auth | CLI dispatch (Gateway is instantiated directly, not via `start --daemon`) |
-| `sources guard-check <url>` | `test_security_download_guard.ts` | `SecurityDownloadGuard.validateUrl/scanFileList` | CLI dispatch + JSON shape |
-| `phase1-check` | (none specifically) | Distribution check in `test_distribution_installability.ts` (different scope) | The CLI handler itself |
-
-Where a row covers multiple commands (e.g., the entire `reversa` group), they share one underlying test family but each individual CLI handler is independently uncovered.
-
----
-
-## Commands with DIRECT COVERAGE
-
-36 commands exercised by subprocess (`spawnSync` / `execSync` on `bin/openlife.js` or `dist/index.js`).
-
-- `ask <mensagem...>` — `test_ask_exit.ts` (exit code & timeout), `test_openlife_gatekeeper_routing.ts` (routing output)
-- `agents list` — `test_openlife_evolution_surface.ts`, `test_cli_doc_parity.ts` (help-only)
-- `agents create <id>` — `test_create_entities.ts`
-- `agents show <id>` — `test_cli_doc_parity.ts`
-- `squads create <id>` — `test_create_entities.ts`
-- `skills list` — `test_openlife_routing_surface.ts`
-- `skills create <id>` — `test_create_entities.ts`
-- `mcp create <id>` — `test_create_entities.ts`
-- `governance policy show` — `test_cli_doc_parity.ts`
-- `route intent <goal...>` — `test_openlife_routing_surface.ts`, `test_openlife_auto_creator_routing.ts`
-- `catalog doctor` (with `--strict`) — `test_openlife_routing_surface.ts`, `test_openlife_auto_creator_routing.ts`, `test_openlife_evolution_surface.ts`
-- `media doctor` — `test_openlife_evolution_surface.ts`
-- `swarm run` — `test_openlife_evolution_surface.ts`
-- `teams upsert <id>` — `test_admin_teams_networks.ts`
-- `networks upsert <id>` — `test_admin_teams_networks.ts`
-- `service create <serviceId>` — `test_openlife_evolution_surface.ts`
-- `service list` — `test_openlife_evolution_surface.ts`
-- `service status <serviceId>` — `test_cli_service_commands.ts`
-- `service pause <serviceId>` — `test_cli_service_commands.ts`
-- `service resume <serviceId>` — `test_cli_service_commands.ts`
-- `service events <serviceId>` — `test_cli_service_commands.ts`
-- `task run <goal...>` — `test_openlife_evolution_surface.ts`
-- `task status <taskId>` — `test_cli_service_commands.ts`
-- `sources scaffold <type> <id>` — `test_sources_scaffold.ts`
-- `sources import-ref <type> <id> <url>` — `test_sources_import_ref.ts`
-- `context init` — `test_enterprise_agentic_core.ts`
-- `context doctor` — `test_enterprise_agentic_core.ts`
-- `context summary` — `test_enterprise_agentic_core.ts`
-- `plugin install <id>` — `test_enterprise_agentic_core.ts`
-- `plugin list` — `test_enterprise_agentic_core.ts`
-- `plugin inspect <id>` — `test_enterprise_agentic_core.ts`
-- `plugin disable <id>` — `test_enterprise_agentic_core.ts`
-- `plugin enable <id>` — `test_enterprise_agentic_core.ts`
-- `pilot create <id>` — `test_enterprise_agentic_core.ts`
-- `pilot list` — `test_enterprise_agentic_core.ts`
-- `pilot report <id>` — `test_enterprise_agentic_core.ts`
-- `pilot complete <id>` — `test_enterprise_agentic_core.ts`
-- `eval judge` — `test_enterprise_agentic_core.ts`
-- `eval report <mission>` — `test_enterprise_agentic_core.ts`
-- `learn from-last-run` — `test_enterprise_agentic_core.ts`
-- `learn list` — `test_enterprise_agentic_core.ts`
-- `enterprise doctor` — `test_enterprise_agentic_core.ts`
-- `aiobuilder validate-catalog` — `test_openlife_evolution_surface.ts`
-- `aiobuilder generate-ui <featureName> --strict-contracts` — `test_reversa_export_and_strict.ts`
-
-(Total above lists 44 lines; the **distinct** count is 36 because some entries cover the same command from two tests, e.g., `agents list` is listed twice.)
-
----
-
-## Test files with no matching CLI command
-
-(These tests cover internal-only behavior. Listing them confirms they were considered and intentionally not mapped to a CLI gap.)
-
-- `test_orchestration_assets_lifecycle.ts` — `SquadAutoCreator`, `SubagentLifecycle` (no direct CLI surface)
-- `test_governance.ts` — `GovernanceLayer` + `GovernanceScopePolicy` unit
-- `test_governance_advanced.ts` — `GovernanceConsentStore`, `SandboxPolicy`, `SecurityEventStore`
-- `test_brain_fallback_chain.ts` / `test_brain_error_diagnostics.ts` — `Brain.think()` internals (no `brain` CLI)
-- `test_phase6_board.ts` / `test_phase6_cadence.ts` / `test_phase6_ops.ts` — `ExecutionBoard`, `RoleHandoff`, `CadenceEngine`, `OperatingSystem`, `ReleaseWorkflow` (Phase-6 ops layer; no CLI)
-- `test_consequence_forecaster.ts` — `ConsequenceForecaster` (no `forecast` CLI)
-- `test_outcome_simulator.ts` — `OutcomeSimulator` (no `simulate` CLI)
-- `test_optimization_loop.ts` — `OptimizationLoop` (no CLI)
-- `test_performance_scorecard.ts` — `PerformanceScorecard` (no CLI)
-- `test_release_gate.ts` — `ReleaseWorkflow` (no CLI; sibling of `governance` story)
-- `test_runtime_policy.ts` — `RuntimePolicy.decide/recordResult/saveSnapshot` (no CLI; surfaces in `models` and `runtime` indirectly)
-- `test_service_completion_policy.ts` — `ServiceCompletionPolicy` (no CLI; gate sits inside `service` runtime)
-- `test_service_guardrails_delete.ts` — `GovernanceLayer.evaluate` again; no CLI
-- `test_telegram_delete_guardrail.ts` — `IntentClassifier` + `Gatekeeper` in-process; no CLI
-- `test_trigger_basic_auth.ts` — Express `/api/v1/trigger` HTTP auth; no CLI
-- `test_daemon_sigterm.ts` — `Gateway.shutdown` SIGTERM behavior; no CLI
-- `test_conversation_memory.ts` — `ConversationMemory` fake-injected
-- `test_memory_promotion.ts`, `test_memory_retention.ts` — Memory layer internals
-- `test_dual_mode.ts` — `ExecutionRouter` mode routing (no `dual-mode` CLI)
-- `test_service_command_surface.ts` — Internal `ServiceStateStore` smoke (already exercised by `test_cli_service_commands.ts` at the CLI layer)
-- `test_distribution_installability.ts` — package layout / file presence; not a command per se
-- `test_openlife_runtime_source_truth.ts` — Source-truth invariant for `AgentRegistry`/`SquadRegistry`
-- `test_openlife_evolution_surface.ts`, `test_openlife_routing_surface.ts`, `test_openlife_auto_creator_routing.ts`, `test_openlife_gatekeeper_routing.ts` — these DO spawn the CLI and are credited above
-
----
-
-## Coverage hot spots and cold spots
-
-- **Hottest group:** `plugin` and `pilot` — 100% direct CLI coverage thanks to `test_enterprise_agentic_core.ts`. The enterprise core is the single best-tested surface in the CLI (covers 18 commands across `plugin`, `context`, `pilot`, `eval`, `learn`, `enterprise`).
-- **Second hottest:** `service` group — 6/6 direct coverage across two tests (`test_cli_service_commands.ts` + `test_openlife_evolution_surface.ts`).
-- **Coldest groups (prioritize for Phase 2/3):**
-  1. **`memory` (5 commands, 0 direct)** — entire surface is uncovered at CLI level even though subsystem tests exist; high user-visibility gap.
-  2. **`models` (6 commands, 0 direct)** — primary cognitive routing UX; mutating `models.json` via CLI is unverified.
-  3. **`benchmark` (3 commands, 0 direct)** — engine is tested but CLI runner/comparer/reporter aren't.
-  4. **`reversa` (6 commands, 0 direct)** — large gap despite extensive subsystem testing; the entire `reversa` CLI surface should be E2E-tested.
-  5. **`designmd` (5 commands, 0 direct)** — same pattern as reversa: subsystem tested, CLI handlers not.
-  6. **`auth` (2 commands, 0 direct)** — OAuth flow is hard to test, but at least a `--help` smoke per command would close the gap and detect regression on lazy-import contract.
-  7. **`smoke` (2 commands, 0 direct)** — by name these are smokes, yet no test exercises them.
-  8. **`orchestration` (2 commands, 0 direct)** — `contract` and `snapshot` outputs are unchecked.
-  9. **Top-level standalone commands**: `restart`, `update`, `up`, `status`, `doctor`, `phase1-check`, `start`, `agent <verb>`, `dream` are all uncovered at the CLI level.
-
-- **Asymmetry to fix:** `agents list/create/show` is fully covered directly, but the parallel `squads list` and `skills list` are not — `test_openlife_evolution_surface.ts` covers `agents list` and `test_openlife_routing_surface.ts` covers `skills list`, but **no test directly exercises `squads list`** (only the subsystem). Trivial to add.
-
-- **Top-level non-coverage is the highest-ROI fix.** A single new `test_top_level_smoke.ts` that spawns each of the 11 top-level standalones with `--help` and the safe ones (`status`, `doctor`, `phase1-check`, `up`) for real would move ~6 commands from "None" to "Direct" with one file.
-
-- **Anomalies in coverage classification:**
-  - `test_cli_doc_parity.ts` is help-only for most groups; counted as direct for `agents show`, `agents list`, `governance policy show` because it actually executes those handlers, but help-only for the rest of its assertions. Documented as Direct where it spawns the handler with real arguments.
-  - `test_openlife_evolution_surface.ts` is the single most-coverage-dense test (it exercises 8 distinct CLI commands across `catalog`, `media`, `agents`, `swarm`, `aiobuilder`, `service`, `task`).
-  - `test_enterprise_agentic_core.ts` covers 18 commands single-handedly.
-  - `test_distribution_installability.ts` covers no CLI command — it only checks package layout (`bin/openlife.js` exists, `scripts/postinstall-check.sh` exists, etc.).
-  - `auth openai` has alias `codex` — both names route to the same handler (line 259); no test covers either name.
-  - `dream` has alias `/dream` — no test covers either alias.

package/docs/audit/integrations/SKIPPED.md

@@ -1,101 +0,0 @@
-# Integrations Skipped — INT-05 / INT-06
-
-**Generated:** 2026-05-11
-**Phase:** 4 — Live Integration Exercise
-
-These integrations were NOT exercised live during Phase 4 because the required credentials or local infrastructure are unavailable in the current environment. Each entry has reproduction steps so anyone with access can run the check later.
-
----
-
-## INT-05 — Ollama (local LLM runtime)
-
-**Status:** SKIPPED — no local Ollama instance running
-**Provider in code:** `Brain.thinkWithOllama()` at `src/orchestrator/Brain.ts`
-**Default URL:** `http://localhost:11434/api/chat` (overridable via `OLLAMA_URL`)
-
-### How to verify once Ollama is available
-
-```bash
-# 1. Install + start Ollama
-curl -fsSL https://ollama.com/install.sh | sh
-ollama pull qwen2.5-coder:7b
-ollama serve  # in another shell
-
-# 2. Verify reachable
-curl -s http://localhost:11434/api/tags | jq
-
-# 3. Write a quick live test (mirror src/test_integration_gemini_live.ts):
-#    - stub modelManager to return { primary: { provider: 'ollama', name: 'qwen2.5-coder:7b', raw: 'ollama/qwen2.5-coder:7b' }, fallbacks: [] }
-#    - call brain.think(...) and assert non-empty response
-
-# 4. Run:
-npm run build && node dist/test_integration_ollama_live.js
-```
-
-**Expected behavior:** `Brain.thinkWithOllama()` should POST to the configured URL and return `data.message.content`.
-
----
-
-## INT-06 — OpenAI / Anthropic / OpenRouter (commercial APIs)
-
-**Status:** SKIPPED — no valid credentials present in `.env`
-**Available providers in code:** `Brain.thinkWithOpenAIAPI()`, `Brain.thinkWithAnthropic()`, `Brain.thinkWithOpenRouter()`
-
-### Current `.env` state (verified Phase 4)
-
-| Variable | Status | Notes |
-|----------|--------|-------|
-| `OPENAI_API_KEY` | Present but invalid | Prefix is `gqwen...` not `sk-...` (per Story 1.3 — credential decision deferred to user) |
-| `ANTHROPIC_API_KEY` | Absent | — |
-| `OPENROUTER_API_KEY` | Absent | — |
-
-### How to verify each once credentials are available
-
-#### OpenAI (real `sk-...` key)
-
-```bash
-# Option A: replace OPENAI_API_KEY in .env with a real sk-... key
-# Option B: set OPENAI_BASE_URL to point to OpenAI-compatible endpoint
-#          (Brain supports this via Story 1.3 OPENAI_BASE_URL infra)
-
-# Then run a live test analogous to gemini live:
-#    stub modelManager → { provider: 'openai-api', name: 'gpt-4o-mini', raw: 'openai-api/gpt-4o-mini' }
-#    assert non-empty response
-```
-
-#### Anthropic
-
-```bash
-# Set ANTHROPIC_API_KEY=sk-ant-... in .env
-# Stub modelManager → { provider: 'anthropic', name: 'claude-3-5-haiku-20241022', raw: 'anthropic/claude-3-5-haiku-20241022' }
-# Run test
-```
-
-#### OpenRouter
-
-```bash
-# Set OPENROUTER_API_KEY=... in .env
-# Stub modelManager → { provider: 'openrouter', name: 'mistralai/mistral-7b-instruct', raw: 'openrouter/mistralai/mistral-7b-instruct' }
-# Run test
-```
-
----
-
-## Why this matters
-
-These gaps don't block the v1.0 milestone — they're orthogonal to the OpenLife framework's correctness. They DO matter for the fallback-chain reliability story (Story 1.3): currently only Gemini is verified live. If Gemini quota is exhausted or the API is down, the documented fallback chain has no live coverage.
-
-**Recommendation for v2:** When credentials become available (for OpenAI/Anthropic/OpenRouter) or local Ollama is provisioned, write the parallel live tests using the same scaffold as `test_integration_gemini_live.ts`. They should each take < 5 minutes to author and < $0.001 to run.
-
----
-
-## Summary
-
-| Integration | Status | Exercised | Log |
-|-------------|--------|-----------|-----|
-| INT-01 Gemini | ✓ Live | Yes | `docs/audit/integrations/gemini.log` |
-| INT-02 Telegram | ✓ Live | Yes | `docs/audit/integrations/telegram.log` |
-| INT-03 HTTP `/api/v1/trigger` | ✓ Live | Yes | `docs/audit/integrations/http-trigger.log` |
-| INT-04 Admin endpoints | ✓ Live | Yes | `docs/audit/integrations/admin.log` |
-| INT-05 Ollama | ⊗ Skipped | No | (this doc) |
-| INT-06 OpenAI/Anthropic/OpenRouter | ⊗ Skipped | No | (this doc) |

package/docs/autonomous-install.md

@@ -1,79 +0,0 @@
-# OpenLife Autonomous Install
-
-Este guia instala o agente autônomo em **local**, **WSL** ou **Linux server** com um comando.
-
-## Comando único mais importante
-
-```bash
-openlife agent install
-```
-
-Depois, para iniciar:
-
-```bash
-openlife agent start
-```
-
-## Pré-requisitos
-
-- Node.js 20+
-- npm 10+
-- Projeto OpenLife clonado
-
-## Instalação rápida
-
-```bash
-./scripts/openlife-autonomous-install.sh
-```
-
-O script executa:
-1. `npm run build`
-2. `openlife agent install`
-3. `openlife mode set autonomous --host claude-code`
-4. Detecta process manager (systemd/supervisor/manual)
-
-## Comando único (CLI)
-
-Você também pode instalar só os artefatos do agente:
-
-```bash
-openlife agent install
-```
-
-Isso cria:
-- `.openlife/agent-runtime.json`
-- `.openlife/agent-queue.json`
-- `.openlife/autonomous-policy.json`
-- `.openlife/autonomous-install-manifest.json`
-- `scripts/openlife-agent-start.sh`
-- `scripts/openlife-agent.service.example`
-- `scripts/openlife-agent.supervisord.conf.example`
-
-## Linux Server com systemd
-
-```bash
-sudo cp scripts/openlife-agent.service.example /etc/systemd/system/openlife-agent.service
-sudo systemctl daemon-reload
-sudo systemctl enable --now openlife-agent
-sudo systemctl status openlife-agent
-```
-
-## Fallback com Supervisor
-
-Se systemd não estiver disponível:
-
-1. Copie `scripts/openlife-agent.supervisord.conf.example` para sua configuração do Supervisor.
-2. Rode:
-
-```bash
-sudo supervisorctl reread
-sudo supervisorctl update
-sudo supervisorctl status
-```
-
-## Execução manual
-
-```bash
-./scripts/openlife-agent-start.sh
-openlife agent status
-```

package/docs/capability-genesis.md

@@ -1,137 +0,0 @@
-# Capability Genesis Engine
-
-> Source: `src/orchestrator/capability/CapabilityGenesisEngine.ts`
-> CLI entry: `openlife create capability "<brief>"` (alias of `openlife aiobuilder create-capability`)
-
-The **Capability Genesis Engine** orchestrates the existing creators
-(`SkillCreator`, `SquadCreator`, `DynamicAgentBuilder`, `WorkflowEngine`)
-to produce a **draft Capability Pack** from a free-text brief.
-
-## Reuse-first contract
-
-Locked decision (2026-05-12): the engine MUST call
-`AssetReuseRouter.find()` before creating anything new.
-
-```
-brief in
-  │
-  ▼
-AssetReuseRouter.find(brief)  ──┐
-  │                              │
-  │                              ▼
-  │                       existing asset(s)?
-  │                              │
-  │  ┌───────────────────────────┤
-  │  │ yes (match >= 80%)        │ no
-  │  ▼                           ▼
-  │  pack.skills.push({          SkillCreator/SquadCreator.create
-  │    source: 'referenced'      under <packDir>/<kind>/<id>/
-  │  })                          pack.skills.push({ source: 'embedded' })
-  └─────────────────────────────┘
-                  │
-                  ▼
-        capability.yaml + INDEX.md
-                  │
-                  ▼
-   CapabilityPackStateStore.init(actor) → draft
-```
-
-The engine never auto-promotes. Operator (or `aiobuilder canonize`)
-moves the pack through the lifecycle.
-
-## Three authoring modes
-
-| Mode | What gets created |
-|---|---|
-| **quick** | 1 skill + manifest only |
-| **professional** | skills + 1 squad |
-| **elite** | skills + squad + workflow stub + quality-gate checklist |
-
-The mode is a *budget*, not a *floor*: if the brief provides hints, the
-engine respects them up to the mode's ceiling.
-
-## Generated structure
-
-For a `--mode elite` run on `"deep research"`:
-
-```
-.catalog/capabilities/deep-research/
-├── capability.yaml
-├── INDEX.md
-├── skills/
-│   ├── source-credibility-scoring.md
-│   ├── evidence-synthesis.md
-│   ├── contradiction-analysis.md
-│   ├── advanced-query-planning.md
-│   └── executive-briefing.md
-├── squads/research-squad/SQUAD.md
-├── workflows/deep-research-flow.yaml
-├── agents/
-├── checklists/
-└── templates/
-```
-
-Each generated file is a **draft stub** with TBD sections — the
-operator refines after generation.
-
-## CLI
-
-```bash
-# Top-level alias (Story 3.2)
-openlife create capability "<brief>" \
-  --id <slug> \
-  --mode quick|professional|elite \
-  --actor <name> \
-  --skills <csv> \
-  --squads <csv> \
-  --workflows <csv>
-
-# Canonical aiobuilder verb (Story 3.3)
-openlife aiobuilder create-capability "<brief>" \
-  --id <slug> \
-  --mode professional \
-  ...
-```
-
-Both invoke the same handler. AIOBUILDER stays canonical.
-
-## Programmatic API
-
-```ts
-import { CapabilityGenesisEngine } from './orchestrator/capability/CapabilityGenesisEngine';
-
-const engine = new CapabilityGenesisEngine();
-const result = engine.generate({
-  description: 'research mid-cap medical devices',
-  packId: 'medical-devices-research',
-  mode: 'professional',
-  actor: 'rafa',
-  hints: {
-    skills: ['source-credibility-scoring', 'evidence-synthesis'],
-    squads: ['medical-research-squad'],
-  },
-});
-
-if (result.ok) {
-  console.log(`Pack at ${result.packDir}`);
-  console.log(`Reused: ${result.reusedAssets.length}`);
-  console.log(`Created: ${result.createdAssets.length}`);
-}
-```
-
-## What's NOT in v1.3
-
-- **Brain-driven proposal** — v1.3 uses a heuristic. v1.4 will wire an
-  LLM call (gated by env) for a richer first draft.
-- **Pack publishing** — exporting a pack to a remote registry. Stays as
-  `SquadCreator.publish()` placeholder.
-- **Pack migration** — schema migrations between pack versions. Stays
-  as `SquadCreator.migrate()` placeholder.
-
-## Test reference
-
-- `src/test_capability_genesis_engine.ts` — 10 scenarios covering
-  generate, reuse, lifecycle promotion gates.
-- `src/test_deep_research_capability.ts` — proves the bundled
-  `deep-research` pack stays intact.
-- `src/test_guided_creator_cli.ts` — covers the CLI surface end-to-end.