ARL Meta-Layer: Observation Layer (Improvement Meta-Process)
HOOTL (Human Out Of The Loop) = human not a synchronous blocking gate during execution. That is the desired outcome of ARL research.
ARL's Observation Layer is the meta-process that sits above the three SDLC layers. Implemented by agentskill-kaizen (post-hoc) and session-historian (transcript data).
Phase Mapping
| Phase |
Skills/Agents |
| Observe |
agentskill-kaizen, session-historian, logging |
| Identify |
hallucination-detector, fact-check, doc-drift-auditor, code-review |
| Accumulate |
knowledge-explorer, refresh-research, research-curator, context-refinement, research-context-agent |
| Improve |
kaizen-improvement, optimize-claude-md, work-backlog-item close, topic-specialist |
ARL Flow
- Observe — session-historian indexes transcripts; agentskill-kaizen analyzes for patterns
- Identify — hallucination-detector (Stop hook), fact-check (VERIFIED/REFUTED), doc-drift-auditor, agentskill-kaizen maps to R1-R10
- Probe — (To be designed.) Triggers: transcript-analysis Dimension 3 (User Frustration), Dimension 9 (Missing Hooks)
- Accumulate — research/ KB, context manifest, Integration Opportunities. Staleness tracking required.
- Improve — Update SAM process, layer docs, skill instructions
The Knowledge Gap
- Human: Expectations, known issues, assumptions often unarticulated
- AI: Massive domain knowledge, holes in specificity/reliability
- Goal: Build knowledge of invisible practices; probe human for specifics
Human-Probing Flow
Design: arl-human-probing-design.md — triggers, probe questions, project-local domain knowledge format, integration with groom-backlog-item and work-backlog-item.
Experiments & Learnings
Flow experiments run in sam-flow-experiments. Learnings from experiments feed back into ARL improvements, skill instructions, and layer docs.
Canonical Provenance
PROVENANCE.md — HOOTL, ARL, agentskill-kaizen; relationship triangle; author design intent.
1---2name: arl-meta-layer-observation-layer-improvement-meta-process3description: Design: arl-human-probing-design.md — triggers, probe questions, project-local domain knowledge format, integration with groom-backlog-item and work-backlog-item.4---5# ARL Meta-Layer: Observation Layer (Improvement Meta-Process)67**HOOTL** (Human Out Of The Loop) = human not a synchronous blocking gate during *execution*. That is the desired outcome of ARL research.89**ARL's Observation Layer** is the meta-process that sits above the three SDLC layers. Implemented by agentskill-kaizen (post-hoc) and session-historian (transcript data).1011---1213## Phase Mapping1415| Phase | Skills/Agents |16|-------|---------------|17| **Observe** | agentskill-kaizen, session-historian, logging |18| **Identify** | hallucination-detector, fact-check, doc-drift-auditor, code-review |19| **Accumulate** | knowledge-explorer, refresh-research, research-curator, context-refinement, research-context-agent |20| **Improve** | kaizen-improvement, optimize-claude-md, work-backlog-item close, topic-specialist |2122---2324## ARL Flow25261. **Observe** — session-historian indexes transcripts; agentskill-kaizen analyzes for patterns272. **Identify** — hallucination-detector (Stop hook), fact-check (VERIFIED/REFUTED), doc-drift-auditor, agentskill-kaizen maps to R1-R10283. **Probe** — (To be designed.) Triggers: transcript-analysis Dimension 3 (User Frustration), Dimension 9 (Missing Hooks)294. **Accumulate** — research/ KB, context manifest, Integration Opportunities. Staleness tracking required.305. **Improve** — Update SAM process, layer docs, skill instructions3132---3334## The Knowledge Gap3536- **Human**: Expectations, known issues, assumptions often unarticulated37- **AI**: Massive domain knowledge, holes in specificity/reliability38- **Goal**: Build knowledge of invisible practices; probe human for specifics3940---4142## Human-Probing Flow4344Design: [arl-human-probing-design.md](./arl-human-probing-design.md) — triggers, probe questions, project-local domain knowledge format, integration with groom-backlog-item and work-backlog-item.4546---4748## Experiments & Learnings4950Flow experiments run in [sam-flow-experiments](https://github.com/Jamie-BitFlight/sam-flow-experiments). Learnings from experiments feed back into ARL improvements, skill instructions, and layer docs.5152---5354## Canonical Provenance5556[PROVENANCE.md](../../../stateless-agent-methodology/research/arl/PROVENANCE.md) — HOOTL, ARL, agentskill-kaizen; relationship triangle; author design intent.
tools-only