Skip to content

BACKLOG.md

Latest commit

 

History

History
130 lines (103 loc) · 18.8 KB

BACKLOG.md

File metadata and controls

130 lines (103 loc) · 18.8 KB
Raw

BACKLOG (prune aggressively; target ≤ 40 items)

Workflow Hook

"Done means" checklist:

  • Update config/tasks/backlog.yaml WI status + evidence pointer (commit/packet)
  • Update LIFEOS_STATE.md (Current Focus/Blockers/Recent Wins)
  • Refresh baseline pack pointer + sha (artifacts/packets/status/Repo_Autonomy_Status_Pack__Main.zip)

Last Updated: 2026-04-27 (rev17)

Architecture Normalization — RESOLVED (per ADR-001 through ADR-004 and changelog A1-A5)

  • Ratify human approval capture contract — ADR-003 ratified 2026-04-26 via PR #38. §9 in COO_Operating_Contract_v1.0.md.
  • Ratify active vs standby COO semantics and sole-writer boundary — ADR-001 ratified 2026-04-24 via PR #36. §7 in COO_Operating_Contract_v1.0.md.
  • Resolve Drive / Workspace role in canon — ADR-004 ratified 2026-04-26 via PR #39. Drive/Workspace = non-canonical collaboration surface.
  • Resolve Hermes ↔ OpenClaw directionality and pushback rules — ADR-002 ratified 2026-04-24 via PR #36. §8 in COO_Operating_Contract_v1.0.md.
  • Reconcile communications draft with canon or mark proposal-only — Changelog A4 implemented 2026-04-27 via PR #40. Communications draft classified proposal-only/non-canonical.
  • Install architecture-maintenance checks after canon is stable — Changelog A5 implemented 2026-04-27 via PR #40. Event-triggered pre-merge architecture impact rule active.

Now (ready soon; not in WIP yet)

P0 (Critical)

  • Run-lock implementation — DoD: Single-flight enforcement, stale lock detection, concurrent test + stale test pass — Owner: antigravity — Context: Build Loop Plan v2.1 Week 1 prerequisite — Done: 2026-02-26 — runtime/orchestration/loop/run_lock.py (142L), wired to spine, 8 tests (test_run_lock.py)
  • Terminal packet + receipt emission — DoD: Schema matches Evidence Contract; dry-run cycle produces valid packet + receipt index — Owner: antigravity — Context: Build Loop Plan v2.1 Week 1 prerequisite — Done: 2026-02-26 — runtime/receipts/invocation_receipt.py (189L), wired to spine, 12 tests (test_invocation_receipt.py, test_invocation_schema.py)
  • Council V2 shadow-mode wiring — DoD: V2 receives payloads in parallel, verdicts logged, does not gate; legacy remains sole gate — Owner: antigravity — Context: Build Loop Plan v2.1 Week 1 prerequisite — Done: 2026-02-27 — ShadowCouncilRunner imported + called in spine._run_chain_steps after review phase (7e36284e); 2 wiring tests added
  • Shadow agent capture mechanism — DoD: Dispatches payload, stores evidence, does not affect pipeline — Owner: antigravity — Context: Build Loop Plan v2.1 Week 1 prerequisite — Done: 2026-02-26 — runtime/agents/shadow_capture.py (185L), wired to spine, 5 tests (test_shadow_capture.py)
  • GitHub Actions feasibility report — DoD: Investigation report at artifacts/reports/github_actions_feasibility.md — Owner: antigravity — Context: Build Loop Plan v2.1 Week 1 (investigation only) — Done: 2026-02-27 — 10 sections, 8 workflows, 5 blockers (B3 RESOLVED), 10 recommendations (87f548c5)
  • Burn-in task curation (CEO approval required) — DoD: 5-7 tasks proposed at artifacts/reports/burn_in_task_proposal.md, CEO approved, ready for Batch 1 — Owner: antigravity — Context: Build Loop Plan v2.1 §1 — Done: 2026-02-27 — Batch 1 completed; report at docs/11_admin/Batch1_BurnIn_Report.md (78473e3)
  • Batch 2 burn-in — DoD: 5 cycles post top-3 fixes; Batch 2 summary report with delta from Batch 1; Council V2 promotion criteria evaluated — Owner: antigravity — Context: Build Loop Plan v2.1 §2 — Preceded by: apply F5 (steward budget), F2 (default allowed_paths), F3 (attempt ledger auto-commit)

P1 (High)

  • Phase 7 prod_ci canonical closure — DoD: default-branch Prod CI Proof workflow dispatches and passes, T-022 repeat-run proof is re-recorded from a clean worktree, and canon is reconciled — Owner: codex — Context: local main is 117 commits ahead of origin/main; gh workflow run .github/workflows/prod_ci_proof.yml --ref main currently returns 404 — Priority: P1
  • Authority audit WP2: CEO decision packet — G-CBS canonicality, Council protocol references, DAP consistency, Build Loop canonicality — Owner: active COO — Context: audit OD-AUDIT-003, C-003, C-008, C-009, C-010 — priority: P1 — See artifacts/plans/WP2_CEO_DECISION_PACKET_2026-04-27.md
  • Authority audit WP3: approval/authority enforcement design — approval_receipt.v1 schema, execution_order.v1 authority fields, parser guard plan — Owner: active COO — Context: audit C-006, C-007, APPROVAL-002 — priority: P1 — See artifacts/plans/WP3_APPROVAL_ENFORCEMENT_DESIGN_2026-04-27.md
  • Authority audit WP4: lifecycle/closure semantics design — succeeded disposition, closed requirements, closure receipt path — Owner: active COO — Context: audit C-016, STATE-002, OD-AUDIT-002 — priority: P1 — See artifacts/plans/WP4_LIFECYCLE_CLOSURE_DESIGN_2026-04-27.md
  • T-027 COO closure packet schemas and loader — DoD: artifacts/coo/schemas.md documents the new packet family; closures.py validates and loads close/context/council-request artifacts; targeted tests pass — Owner: codex — Context: first slice of the COO coordination loop; file-based artifacts only — Priority: P1
  • T-028 decision-support backlog field and dispatch gate — DoD: decision_support_required round-trips through backlog state transitions and blocks L0 auto-dispatch until the latest matching council request is resolved — Owner: codex — Context: second slice of the COO coordination loop — Priority: P1
  • T-029 coo process-closures and orientation updates — DoD: python3 -m runtime.cli coo process-closures validates and summarizes closure artifacts; COO brief now reads closures and artifacts/for_ceo/ — Owner: codex — Context: third slice of the COO coordination loop — Priority: P1
  • T-030 sprint-close emission and CT-6 governance hook — DoD: wrappers emit sprint_close_packet.v1; governance and handoff updates land only after Council approval — Owner: codex — Context: final protected slice of the COO coordination loop; blocked on Council ruling — Priority: P1
  • Ledger Hash Chain (Trusted Builder P1) — DoD: Tamper-proof linking of bypass records — Owner: antigravity — Context: Deferred from Trusted Builder v1.1 Ratification
  • OpenClaw distill lane operational rollout — DoD: Session-scoped shadow rollout completed with 12-run / 2-session clean window, CEO-approved shadow success receipt recorded, models status active promotion gated by fresh preflight + forced-failure drill + CEO approval, runbook/state updated — Owner: Substrate — Context: Collaborators COO + CEO; distill lane is merged on main but not yet an active operating practice; milestone 1 keeps raw path authoritative and status --all --usage shadow-only — Priority: P1
  • Bypass Monitoring (Trusted Builder P1) — DoD: Alerting on high bypass utilization — Owner: antigravity — Context: Deferred from Trusted Builder v1.1 Ratification
  • Semantic Guardrails (Trusted Builder P1) — DoD: Heuristics for meaningful changes — Owner: antigravity — Context: Deferred from Trusted Builder v1.1 Ratification
  • Fix test_steward_runner.py (25/27 failing) — DoD: Tests pass or are properly restructured — Owner: antigravity — Context: Import/fixture issues, not code bugs; currently skipped on WSL (build/doc-refresh-and-test-debt)
  • Finalize Intent_Routing_Rule v1.1 — DoD: Markers removed — Owner: antigravity
  • Finalize Test_Protocol v2.0 — DoD: Markers removed — Owner: antigravity
  • Finalize Tier_Definition_Spec v1.1 — DoD: Markers removed — Owner: antigravity
  • Finalize ARTEFACT_INDEX_SCHEMA v1.0 — DoD: Markers removed — Owner: antigravity
  • Finalize QUICKSTART v1.0 — DoD: Context scan pass complete — Owner: antigravity
  • Fix claude_doc_stewardship_gate.py INDEX.md timestamp bug — DoD: --auto-fix handles [Last Updated: YYYY-MM-DD (revN)] format correctly — Owner: antigravity — Context: Auto-fix regex corrupts INDEX.md timestamp in bracket format (P2)

Next (valuable, but not imminent)

Agent Efficiency Initiative

See living doc: artifacts/plans/2026-04-06-agent-efficiency-improvements.md

  • T-AE-01 Run-level token ledger + CLI estimation — DoD: Cumulative token spend visible across all agent calls in a run; 80% budget warning emitted before exhaustion; CLI invocations contribute estimated_tokens (not just API calls) — Owner: Substrate — Context: Port aggregation pattern from autonomous_build_cycle.py:337 into LoopSpine/dispatch/COO path; extend budgets.py:23 and api.py:671 — Priority: P1 — Done: 2026-04-07 — Evidence: runtime/agents/api.py, runtime/orchestration/loop/budgets.py, runtime/orchestration/loop/spine.py; pytest runtime/tests -q PASS (3058 passed, 6 skipped); focused AE token suite PASS (24 passed)

  • T-AE-02 COO context telemetry + stable block reuse — DoD: build_propose_context() logs serialized byte sizes per block; stable blocks (REPO_MAP, delegation_envelope) get content-addressed IDs excluding ephemeral fields; repeated same-session invocations skip re-serializing unchanged blocks — Owner: Substrate — Context: context.py:148 (cache keys), context.py:152 (slim actionable_tasks fields), invoke.py:105 — Priority: P1 — Done: 2026-04-07 — Evidence: runtime/orchestration/coo/context.py; runtime/tests/orchestration/coo/test_context.py; pytest runtime/tests -q PASS (3058 passed, 6 skipped); focused AE token suite PASS (24 passed)

  • T-AE-03 REPO_MAP freshness gate in start_build.py — DoD: start_build.py warns (or blocks, configurable) when REPO_MAP.md is >7 days old; reuses existing detector at scripts/entropy/scan_v1.py:55; no auto-commit automation — Owner: Substrate — Context: REPO_MAP was 25 days stale at 2026-04-06 analysis — Priority: P1 — Done: 2026-04-07 — Evidence: scripts/workflow/start_build.py, scripts/entropy/scan_v1.py, runtime/tests/test_build_start_close.py; pytest runtime/tests -q PASS (3058 passed, 6 skipped); focused AE token suite PASS (24 passed)

  • Tech Debt Audit Follow-up — DoD: Address items in docs/11_admin/TECH_DEBT_INVENTORY.md as their trigger conditions are met — Context: Inventory created 2026-02-27 during 3-pass audit; 5 resolved items, 8 tracked items with explicit triggers — Priority: P2

  • Strategic refresh of lifeos-master-operating-manual to v2.2 — Scope: Strategic rewrite and alignment with current runtime state — Deliverable: Diff-based update reflecting consolidation, alignment, and automation changes — Owner: antigravity — Context: Out-of-band for admin consolidation initiative; deferred

  • Configure GCP Employee instance — Context: Hardened OpenClaw config, dedicated accounts, Tailscale

  • Revenue Track: LinkedIn daily posts — Context: COO drafts, CEO reviews. Seed: "What autonomous AI agents actually cost to run"

  • Revenue Track: B5 Governance Guide — Context: 19,500 lines of real code + 1,440 tests backing the content

  • Mission Type Extensions — Why Next: Add new mission types based on backlog needs

  • Gate 6: Agent-Agnostic Gate Runner — DoD: Refactor claude-specific gates into coo gate run-all; wire into coo land — Owner: antigravity — Context: Doc stewardship status report identifies Claude-specific gates as non-portable

  • Tech Debt: Rehabilitate Legacy Git Workflow Tests — Context: Quarantined to archive_legacy_r6x due to missing run_cmd mock. Rehabilitate or remove.

COO Step 6 Known Gaps (carried forward 2026-03-08)

  • COO Sandboxing Decision — DoD: Council ruling on OS-level containment strategy — Owner: Council — Context: COO currently runs with full filesystem + exec access (sandboxed: false); autonomy boundary is delegation envelope + fail-closed reasoning only. Requires architectural decision before L1/L2 autonomy promotion — Priority: P2 — Done: 2026-03-24 — Council ruling APPROVED (2026-03-20); COO promoted to unsandboxed prod L3; manifest active; Gate-6 UAT 5/5 PASS
  • COO schema drift guard — DoD: output_schema in context.py and artifacts/coo/schemas.md stay in sync; add CI check or single source-of-truth — Owner: Substrate — Context: Step 6 gap #3; build_propose_context() embeds schema inline, divergence from schemas.md is silent — Priority: P2
  • cmd_coo_direct() live Stage A parity — DoD: Escalation and ambiguous parity cases pass live COO replay (not just mock tests) — Owner: Substrate — Context: Step 6 gap #4; only mocked tests exist; Stage A escalation/ambiguous cases were SKIPPED — Priority: P2 — Done: 2026-04-02 — T-016 closed from Gate 6 evidence plus direct-path coverage on main
  • COO retry/backoff in invoke_coo_reasoning() — DoD: Gateway timeouts trigger exponential backoff (max 2 retries) before raising InvocationError — Owner: Substrate — Context: Step 6 gap #5; current adapter is single-shot; gateway restarts cause cascading failures — Priority: P3 — Done: 2026-03-27 — merged 1cd86177 + 69eb7a10
  • COO cron/event trigger — DoD: lifeos coo propose runs on a schedule or backlog-change event without manual invocation — Owner: Wiring — Context: Step 6 gap #6; currently manual-pull only — Priority: P2 — Done: 2026-03-09 — OpenClaw cron heartbeat registered (T-018)
  • coo.md output schema section — DoD: config/agent_roles/coo.md includes inline output schema matching context.py; schemas.md cross-referenced — Owner: Docs — Context: Step 6 gap #7; schema lives only in schemas.md + context.py, not in the role definition the COO reads — Priority: P2 — Done: 2026-04-02 — T-019 closed with prompt authority split in config/agent_roles/coo.md and config/coo/prompt_canonical.md
  • _normalize_proposal_indentation() field coverage — DoD: Either generalize the normalizer or add a regression test that fails when new COO sub-keys appear at column 0 — Owner: Substrate — Context: Step 6 gap #2; currently hard-codes 4 field names; new sub-keys silently ignored — Priority: P3 — Done: 2026-04-02 — T-020 closed with fail-closed normalization coverage in main
  • Thread OPENCLAW_BIN into promotion state version check — DoD: _current_openclaw_version() in openclaw_promotion_state.py uses OPENCLAW_BIN env var (already set by openclaw_coo_update_protocol.sh) instead of bare PATH lookup; test with fake binary at explicit path — Owner: Substrate — Context: Codex review finding on build/openclaw-runtime-upgrade; version check gates promotion but resolves binary from untrusted PATH; defense-in-depth, not urgent since orchestrated-mode + attestation TTL bound the attack window — Priority: P2

Later (not actionable / unclear / exploratory)

  • Fuel track exploration — Why Later: Not blocking Core; future consideration per roadmap
  • Productisation of Tier-1/Tier-2 engine — Why Later: Depends on Core stabilisation

Done (last ~25 only)

  • W5-T02 Checkpoint/Resume E2E Proof — Date: 2026-02-19 — 6 integration tests proving full checkpoint/resume cycle with policy hash continuity and ledger anchoring. Evidence: artifacts/evidence/W5_T02_checkpoint_resume_proof.txt
  • W4-T03 Worktree Dispatch Governance — Date: 2026-02-18 — Spine now enforces isolated worktree lifecycle + clean-worktree fail-closed check (runtime/orchestration/loop/spine.py)
  • W4-T04 Validator Lifecycle Hooks (OpenClaw Path) — Date: 2026-02-18 — OpenClaw execution bridge + CLI command + lifecycle-integrated test coverage (runtime/orchestration/openclaw_bridge.py, runtime/cli.py)
  • Guard deprecated autonomous_build_cycle path (W0-T04) — Date: 2026-02-13 — cli.py deprecation guard + autonomous_build_cycle.py migration notice
  • Doc freshness skeleton gate (W0-T06) — Date: 2026-02-14 — Runtime status generator, checkpoint artifacts, close-build integration
  • Fix test_e2e_smoke_timeout.py (W0-T05) — Date: 2026-02-14 — Tests already passing, no work needed
  • E2E Loop Test: Real task through full pipeline — Date: 2026-02-14 — run_20260214_053357 completed full 6-phase chain, finalized Emergency_Declaration_Protocol v1.0, discovered/fixed model config issues
  • Finalize Emergency_Declaration_Protocol v1.0 — Date: 2026-02-14 — Completed via autonomous spine run, markers removed, status changed to ACTIVE
  • EOL Clean Invariant Hardening — Date: 2026-02-10 — 289-file renormalization, config-aware clean gate, acceptance closure validator, 37 tests, EOL_Policy_v1.0
  • Install OpenClaw COO on WSL2 — Date: 2026-02-11 — OpenClaw installed and acceptance-verified
  • Manual v2.1 Reconciliation Sprint — Date: 2026-02-08 — CRLF fix, 36 tests re-enabled, free Zen models, manual corrected
  • Deletion Safety Hardening (Article XIX) — Date: 2026-02-08
  • Documentation Stewardship — Date: 2026-02-08 — 5 root docs relocated to canonical locations
  • StewardMission Git Ops (Full Implementation) — Date: 2026-02-08 — 691 lines, real git ops, governance guards
  • LLM Backend Configuration — Date: 2026-02-08 — config/models.yaml with 5 agents, fallback chains
  • Phase 4 (4A0-4D) Full Stack Merge — Date: 2026-02-03 — Autonomous build loop canonical
  • Repository Branch Cleanup — Date: 2026-02-03 — 9 branches assessed, 8 archived, single canonical main
  • E2E Test: Runtime Greet — Date: 2026-02-01
  • Complete Deferred Evidence: F3/F4/F7 — Date: 2026-01-27
  • Standardize Raw Capture Primitive — Date: 2026-01-18
  • Finalize CSO_Role_Constitution v1.0 — Date: 2026-01-23
  • Git Workflow Protocol v1.1 Impact — Date: 2026-01-16
  • Grok Fallback Debug & Robustness Fixes v1.0 — Date: 2026-01-18
  • CLI & Mission Hardening v1.0 — Date: 2026-01-13
  • Tier-3 CLI Integration (Full) — Date: 2026-01-13