Orchestrata Ultimate › Plans › research › audit-05-spec-gaps
Audit 05 — Spec-Compliance Gap Analysis (000–105)
Source: plans/research/audit-05-spec-gaps.md at 8a2cb42ad · last changed 2026-07-27 · first committed 2026-07-04 · 40 lines.
Prose pending.
Plan text
Audit 05 — Spec-Compliance Gap Analysis (000–105)
Method caveats
- Placeholder markers BANNED + enforced (
review_checks.py,core/standards.pyNO_PLACEHOLDER). Gaps = absent modules, not TODOs. Only 1 real TODO:gateway/test_runner.py:288(coverage). - Per-spec
compliance-audit.md"Missing" flag UNRELIABLE (false-flags real modules by class-name matching). specs/SPEC_INVENTORY.mdstale (2026-04-08, ~24 of 61).plans/rtm.md: 40.4% coverage, 129 verified / 190 missing / 3197 orphan.
Inventory: 61 dirs, duplicate 007 (prolog-governance + verification)
Deferred-item verification
- 103 tool-calling-refactor — NO LONGER deferred; IMPLEMENTED + wired (
refactor/agent.py,orchestrator/intent_broker.py:309-320 _handle_refactor, gatedenable_tool_refactor=True). Thin (DeepSeek only). - 104 openspec-format — IMPLEMENTED (small,
openspec/484 LOC + tests). - 017 calendar — STILL DEFERRED: Python CalDAVProvider/GoogleCalendarProvider/SessionCalendarSync ABSENT; JS gadget only.
Core areas SOLID (real + tested)
volks (agents/ 81 files/24.5k), postman (31/13.2k), cxp (38/10.9k), knowledge_graph (39/11.3k), session (59/13.3k), gateway (31/7.3k), governance/PAM (24/6.3k + prolog), harness, memory (hot/cold/decay), goal.
NB: "solid" = code exists + tested. The volks/comms/context/tracing/KG audits found the CAPABILITIES are largely UNWIRED within these solid modules (dead personas, unused tool loop, orphaned tracer, unwired injector, no volk↔volk mesh, no live status). Mandate value = wiring + semantic enrichment + tool suite, not rewrites.
Prioritized gap table
| # | Spec | Status | Missing | Effort |
|---|---|---|---|---|
| 1 | 063–072 tool suite (10 specs) | DOC-ONLY, zero code | No REPL/LSP/PowerShell/RemoteTrigger/Cron/Task/Team/MCP/ToolSearch/Worktree classes exist. Only 7 tools in agents/tool_registry.py. "42-tool parity" unbacked. |
XL |
| 2 | 105 ontology | PARTIAL | domaingen/ real (682) but stardog.py live-only, no CI-verifiable persistence |
M |
| 3 | 103 refactor | IMPL-thin | single driver, small tests, no rollback/verify depth | M |
| 4 | 052 todoist | UI-only | no Python pkg (exec-plan Step 222 never run); JS gadget only | S-M |
| 5 | 017 calendar | UI-only/deferred | Python providers absent | M |
| 6 | 021 opencode | DOC-ONLY | no Python (external/aspirational?) | M/WONTFIX |
| 7 | 022 tui | STUB | no textual/rich TUI; ultimatetui.py is a bridge to CheetahClaws |
M/reconcile |
| 8 | 104 openspec | IMPL-thin | shallow parse/render | S |
| 9 | 000 cli | IMPL, status-drift | code real, spec says "Design Phase" | XS doc |
Cross-cutting
- RTM: 40.4%, 190 missing, 3197 orphan. Close = L.
- Duplicate spec 007 (numbering/RTM hazard). XS.
- Stale
SPEC_INVENTORY.mdasserts "100% tool parity" contradicted by gap #1. XS. plans/backlog-plan.mdalready cleared (session-hook chores, not spec gaps).- exec-plan Step 221 (real CheetahClaws TUI), Step 222 (real Todoist) never executed → gaps #4/#7.
Bottom line
Core loop (Cx→…→Cd→Qa→Cp) real. Biggest doc-vs-code gap = 063–072 tool suite (zero code). Secondary = thin edges of 103/105 + UI-only 052/017.