research(companion-reuse): Sub-wave A — aif-handoff autoQueueMode deep-dive

[artyhoo]

Summary

• Sub-wave A of umbrella companion-reuse-deep-dive: re-evaluates autoQueueMode LOGIC as standalone sub-component, separate from full Kanban runtime (SSOT docs: post-1A follow-up (Class C reviewer-discipline, Q3/Q5, README absolutism) #67, REJECT — preserved as-is).
• Verdict: ADOPT VOCABULARY (primary, tag schema roadmapAlias/phase:N/seq:NN) + REFERENCE (pool-slot-allocation algorithm as design precedent). Match score revised from ~22% (muddied-scope full-runtime) to ~45% (corrected sub-component scope).
• Corrects T16 muddied-scope trap (T-CR-A: «multi-project across repos» framing vs «within-one-project, multi-initiative») from 2026-05-23 initial sweep.
• SSOT new row proposed at [next-available-slot] (parallel sub-waves dispatched simultaneously; maintainer resolves number at merge).

§1.7 Forward-check applied

• build-first-reuse-default.md §3 6-layer search: 4 DeepWiki probes (within-one-project disambiguated) + WebSearch + SSOT consult (docs: post-1A follow-up (Class C reviewer-discipline, Q3/Q5, README absolutism) #67 + 8 other aif-handoff rows). No negative-existence claim.
• ai-laziness-traps.md §2 T16: explicit upstream vs our problem-class walk in §3 match table (9 dimensions scored).
• ai-laziness-traps.md §2 T20: verdict backed by DeepWiki probe outputs + SSOT cross-reference + falsifier in same section.
• no-paid-llm-in-ci.md §1: all research via DeepWiki MCP (free) + WebSearch. No API-billed calls.
• doc-authority-hierarchy.md §2-§3: patch carries Status/Date/Authoritative-for/NOT-authoritative-for header.
• See patch §6 for full forward-check.

§1.7 Backward-check applied

• SSOT docs: post-1A follow-up (Class C reviewer-discipline, Q3/Q5, README absolutism) #67 REJECT verdict preserved as-is. New row is a separate entry; additive note proposed on docs: post-1A follow-up (Class C reviewer-discipline, Q3/Q5, README absolutism) #67 «Trigger to revisit» only.
• No contradiction with prior R-patches (2026-05-23 meta-orchestrator, 2026-05-26 mode-triage, stage-5-dogfood).
• Sub-waves B/C/D: different companion projects, no overlap.
• See patch §7 for full backward-check.

How to verify

• Patch file present at docs/meta-factory/research-patches/2026-05-26-companion-reuse-aif-handoff-autoqueue.md — verified: file exists with <!-- scope:companion-reuse-aif-handoff-autoqueue --> on line 1, §0–§8 complete
• DeepWiki sweep ≥3 phrasings all with within-one-project disambig — verified: 4 probes, all T-CR-A compliant, URLs cited in patch §2
• Verdict cites SSOT row + DeepWiki output + falsifier — verified: patch §4 last paragraph
• All 19 principle tests pass — verified: pre-push output shows 130 passed, 4 skipped, 0 failed
• markdownlint clean — verified: pre-push output shows 0 errors
• Cold-QA pass performed pre-PR per T19 — 2 cold-QA findings caught and fixed (bare URLs MD034, fenced blocks without language MD040, missing scope annotation)
• Maintainer reviews verdict + SSOT new-row proposal — owner: maintainer
• Maintainer resolves [next-available-slot] placeholder at merge — owner: maintainer
• Maintainer applies additive note to SSOT docs: post-1A follow-up (Class C reviewer-discipline, Q3/Q5, README absolutism) #67 «Trigger to revisit» if accepted — owner: maintainer

Scope-out

• I-phase implementation of any verdict (this is research-only).
• Substrate edits (out of scope per kickoff §3).
• New principle tests (out of scope per kickoff §3).
• Amending SSOT docs: post-1A follow-up (Class C reviewer-discipline, Q3/Q5, README absolutism) #67 verdict (preserved as-is; only additive note on «Trigger to revisit» proposed).

[artyhoo]

Review finding — orchestrator self-audit (post-aggregate, 2026-05-26)

MAJOR: «zero-infra» axis in SSOT #67's REJECT rationale weakens under satellite-positioning

Per README.md:72-74 the project's positioning is satellite of aif-handoff (literally: «Is: companion to AI Factory + aif-handoff + Superpowers» / «Isn't: task orchestration / swarm coordination (use aif-handoff)»).

SSOT #67 rejection cites three axes (line 135):

• §7.9 (must NOT write code) vs aif-handoff writes code ← still load-bearing
• §7.10 (zero infra) vs Docker+PG requirement ← weakens under satellite-positioning because the consumer profile «uses aif-handoff» already has Docker+PG; we don't add infra, we ride on top
• §7.1 (user-invoked slash) vs cron-driven ← still load-bearing for UX (user-invoked entry)

The autoQueueMode LOGIC sub-component evaluation in this Sub-wave A is correct under «zero-infra-on-our-side» interpretation. But the broader SSOT #67 REJECT verdict for the full Kanban runtime warrants re-evaluation under satellite-positioning: a CC-native slash-command wrapper around aif-handoff Kanban (passing Docker+PG burden to the satellite-host) might re-classify from REJECT to ADAPT.

This is out of scope for Sub-wave A's amendment (which keeps #67 REJECT preserved as-is per kickoff §3 «No re-litigating R-phase verdicts that already shipped to staging — only AMENDMENTS with corrected match-scores»). But it's a real finding for a separate future R-phase umbrella «#67-reevaluation-under-satellite-positioning» — surfaced here for visibility, not actionable in this PR.

Optional: AI-agnostic vs CC-session-bound framing

Per README.md:65+README.md:74, the project is AI-agnostic (Claude Code, Cursor, Cline, Codex, Aider as deployment surfaces; agents under agents/ AI-agnostic). Some places in the patch / orchestrator aggregate framed our constraint as «CC session-bound» — the more precise constraint is «session-bound in any of 5 supported AI harnesses, user-invoked entry point, not cron». The structural mismatch with aif-handoff's pollAndProcess cron remains, but the rationale should reflect AI-agnostic positioning. No change required in this PR; flagged for future patches.

---

Orchestrator self-audit (T19 own-cold-QA-before-handoff applied to aggregate). Sub-wave A's autoQueueMode-specific evaluation stands; broader #67 re-evaluation tracked as separate future umbrella.