@jaimevalasek/aioson 1.8.0 → 1.9.1

package/template/.aioson/docs/LAYERS.md

@@ -1,89 +1,89 @@
----
-description: "Guide to the three project memory layers: rules, docs, and design-docs — when to use each"
-agents: []
----
-
-# AIOSON Project Memory Layers
-
-Four directories accumulate project knowledge over time.
-Each has a different purpose and a different update cadence.
-
----
-
-## Layer 1 — `.aioson/rules/`
-
-**What it is:** behavioral overrides for agents.
-**Who writes it:** the user, or promoted from recurring @dev patterns.
-**When to use:** when you want to enforce a convention that overrides agent defaults — globally or for specific agents.
-**Cadence:** stable. Rules change rarely; only when a convention is proven wrong or superseded.
-
-Examples of good rules:
-- "All API routes must follow REST naming conventions in this project"
-- "Never use float for monetary values — use integer cents"
-- "The @dev agent must always write a migration for schema changes"
-
-See `rules/README.md` for format and frontmatter reference.
-
----
-
-## Layer 2 — `.aioson/docs/`
-
-**What it is:** domain knowledge and technical reference that agents load on demand.
-**Who writes it:** the user or @architect, based on real integration and domain complexity.
-**When to use:** when multiple agents across different features need the same external context — API behavior, third-party quirks, data model explanations, integration patterns.
-**Cadence:** updated when the referenced system changes, not after every feature.
-
-Examples of good docs:
-- `stripe-integration-context.md` — describes webhook event model, idempotency keys used
-- `auth-rbac-model.md` — explains the role/permission system as it stands in production
-- `legacy-api-behavior.md` — documents known quirks of an external API affecting multiple features
-
-See `docs/README.md` for format and naming conventions.
-
----
-
-## Layer 3 — `.aioson/design-docs/`
-
-**What it is:** structural code governance: folder structure, componentization, reuse, naming, and file-size thresholds.
-**Who writes it:** installed by AIOSON, then edited by the project team when conventions change.
-**When to use:** before architectural structure decisions and before implementation that creates files, splits modules, introduces reusable code, or names APIs.
-**Cadence:** stable. These files are project-local and preserved on update.
-
----
-
-## Layer 4 — `.aioson/context/design-doc*.md`
-
-**What it is:** living decision document for the current feature or project scope.
-**Who writes it:** @discovery-design-doc.
-**Who updates it:** @dev at feature close, @discovery-design-doc when resuming.
-**When to use:** automatically — one per feature (`design-doc-{slug}.md`) or one project-wide (`design-doc.md`).
-**Cadence:** updated at the end of each feature implementation session. Decisions are append-only — never deleted.
-
----
-
-## Decision Guide
-
-| Situation | Where it goes |
-|-----------|--------------|
-| Enforce a coding convention for this project | `rules/` |
-| Agents must always know about an external API behavior | `docs/` |
-| Enforce structural code quality guidance | `design-docs/` |
-| Document the scope and decisions for a specific feature | `design-doc-{slug}.md` |
-| Log a global project-wide architecture decision | `design-doc.md` |
-| Promote a recurring @dev pattern | `rules/` via @dev promotion |
-| Document an integration used by 3+ features | `docs/` |
-| Record what was decided during a single feature session | `design-doc-{slug}.md` (Decisions already made section) |
-
----
-
-## What NOT to put in these layers
-
-| Content | Where it actually belongs |
-|---------|--------------------------|
-| Feature requirements | `requirements-{slug}.md` |
-| PRD / product scope | `prd-{slug}.md` |
-| Execution sequence | `implementation-plan-{slug}.md` |
-| Current implementation state | `spec-{slug}.md` |
-| Project-wide context | `project.context.md` |
-| Domain entity map | `discovery.md` |
-| Technical architecture | `architecture.md` |
+---
+description: "Guide to the three project memory layers: rules, docs, and design-docs — when to use each"
+agents: []
+---
+
+# AIOSON Project Memory Layers
+
+Four directories accumulate project knowledge over time.
+Each has a different purpose and a different update cadence.
+
+---
+
+## Layer 1 — `.aioson/rules/`
+
+**What it is:** behavioral overrides for agents.
+**Who writes it:** the user, or promoted from recurring @dev patterns.
+**When to use:** when you want to enforce a convention that overrides agent defaults — globally or for specific agents.
+**Cadence:** stable. Rules change rarely; only when a convention is proven wrong or superseded.
+
+Examples of good rules:
+- "All API routes must follow REST naming conventions in this project"
+- "Never use float for monetary values — use integer cents"
+- "The @dev agent must always write a migration for schema changes"
+
+See `rules/README.md` for format and frontmatter reference.
+
+---
+
+## Layer 2 — `.aioson/docs/`
+
+**What it is:** domain knowledge and technical reference that agents load on demand.
+**Who writes it:** the user or @architect, based on real integration and domain complexity.
+**When to use:** when multiple agents across different features need the same external context — API behavior, third-party quirks, data model explanations, integration patterns.
+**Cadence:** updated when the referenced system changes, not after every feature.
+
+Examples of good docs:
+- `stripe-integration-context.md` — describes webhook event model, idempotency keys used
+- `auth-rbac-model.md` — explains the role/permission system as it stands in production
+- `legacy-api-behavior.md` — documents known quirks of an external API affecting multiple features
+
+See `docs/README.md` for format and naming conventions.
+
+---
+
+## Layer 3 — `.aioson/design-docs/`
+
+**What it is:** structural code governance: folder structure, componentization, reuse, naming, and file-size thresholds.
+**Who writes it:** installed by AIOSON, then edited by the project team when conventions change.
+**When to use:** before architectural structure decisions and before implementation that creates files, splits modules, introduces reusable code, or names APIs.
+**Cadence:** stable. These files are project-local and preserved on update.
+
+---
+
+## Layer 4 — `.aioson/context/design-doc*.md`
+
+**What it is:** living decision document for the current feature or project scope.
+**Who writes it:** @discovery-design-doc.
+**Who updates it:** @dev at feature close, @discovery-design-doc when resuming.
+**When to use:** automatically — one per feature (`design-doc-{slug}.md`) or one project-wide (`design-doc.md`).
+**Cadence:** updated at the end of each feature implementation session. Decisions are append-only — never deleted.
+
+---
+
+## Decision Guide
+
+| Situation | Where it goes |
+|-----------|--------------|
+| Enforce a coding convention for this project | `rules/` |
+| Agents must always know about an external API behavior | `docs/` |
+| Enforce structural code quality guidance | `design-docs/` |
+| Document the scope and decisions for a specific feature | `design-doc-{slug}.md` |
+| Log a global project-wide architecture decision | `design-doc.md` |
+| Promote a recurring @dev pattern | `rules/` via @dev promotion |
+| Document an integration used by 3+ features | `docs/` |
+| Record what was decided during a single feature session | `design-doc-{slug}.md` (Decisions already made section) |
+
+---
+
+## What NOT to put in these layers
+
+| Content | Where it actually belongs |
+|---------|--------------------------|
+| Feature requirements | `requirements-{slug}.md` |
+| PRD / product scope | `prd-{slug}.md` |
+| Execution sequence | `implementation-plan-{slug}.md` |
+| Current implementation state | `spec-{slug}.md` |
+| Project-wide context | `project.context.md` |
+| Domain entity map | `discovery.md` |
+| Technical architecture | `architecture.md` |

package/template/.aioson/docs/README.md

@@ -1,76 +1,76 @@
----
-description: "Guide to the docs layer: domain knowledge and technical reference for agents"
-agents: []
----
-
-# Project Docs
-
-Files in this directory are domain knowledge and technical reference that agents load on demand.
-
-Unlike `rules/` (which enforce conventions), docs explain **how something works** — external APIs, data models, integration patterns, legacy behavior, third-party quirks.
-
----
-
-## Frontmatter Format
-
-```yaml
----
-description: "Short description of what this doc covers — used by agents to decide relevance"
-scope: "global"       # or a feature slug if doc is scoped to one area
-agents: []            # empty = any agent may load; or restrict: [dev, architect]
----
-```
-
----
-
-## Field Reference
-
-| Field | Required | Description |
-|-------|----------|-------------|
-| `description` | yes | What this doc covers — agents use this to decide whether to load it |
-| `scope` | no | `global` (default) or a feature slug for narrow scope |
-| `agents` | no | If empty, any agent may load; if listed, only those agents load it |
-
----
-
-## Loading Behavior
-
-Agents load docs when the `description` frontmatter is relevant to the current task.
-An empty `agents:` field means any agent can load the doc.
-Docs are loaded in addition to the current context — not as replacements.
-
----
-
-## When to Create a Doc
-
-Create a doc when:
-- Two or more features need the same external context
-- An integration has non-obvious behavior that has caused bugs or would cause rework
-- A data model exists only in production and is not derivable from the codebase
-- A third-party API has quirks that affect how agents implement against it
-
-Do NOT create a doc for:
-- Feature-specific decisions (use `design-doc-{slug}.md` instead)
-- Project conventions (use `rules/` instead)
-- Current implementation state (use `spec-{slug}.md` instead)
-
----
-
-## Naming Convention
-
-Use kebab-case. Name the file after the system or concept, not the feature:
-
-| Good | Bad |
-|------|-----|
-| `stripe-webhook-behavior.md` | `billing-feature-notes.md` |
-| `auth-rbac-model.md` | `auth-stuff.md` |
-| `legacy-api-quirks.md` | `misc-notes.md` |
-| `email-delivery-constraints.md` | `sendgrid.md` (too generic) |
-
----
-
-## Example
-
-See `example-external-api-context.md` in this directory for a working template.
-
-For the layer separation guide (when to use `docs/` vs `rules/` vs `design-doc`), see `LAYERS.md`.
+---
+description: "Guide to the docs layer: domain knowledge and technical reference for agents"
+agents: []
+---
+
+# Project Docs
+
+Files in this directory are domain knowledge and technical reference that agents load on demand.
+
+Unlike `rules/` (which enforce conventions), docs explain **how something works** — external APIs, data models, integration patterns, legacy behavior, third-party quirks.
+
+---
+
+## Frontmatter Format
+
+```yaml
+---
+description: "Short description of what this doc covers — used by agents to decide relevance"
+scope: "global"       # or a feature slug if doc is scoped to one area
+agents: []            # empty = any agent may load; or restrict: [dev, architect]
+---
+```
+
+---
+
+## Field Reference
+
+| Field | Required | Description |
+|-------|----------|-------------|
+| `description` | yes | What this doc covers — agents use this to decide whether to load it |
+| `scope` | no | `global` (default) or a feature slug for narrow scope |
+| `agents` | no | If empty, any agent may load; if listed, only those agents load it |
+
+---
+
+## Loading Behavior
+
+Agents load docs when the `description` frontmatter is relevant to the current task.
+An empty `agents:` field means any agent can load the doc.
+Docs are loaded in addition to the current context — not as replacements.
+
+---
+
+## When to Create a Doc
+
+Create a doc when:
+- Two or more features need the same external context
+- An integration has non-obvious behavior that has caused bugs or would cause rework
+- A data model exists only in production and is not derivable from the codebase
+- A third-party API has quirks that affect how agents implement against it
+
+Do NOT create a doc for:
+- Feature-specific decisions (use `design-doc-{slug}.md` instead)
+- Project conventions (use `rules/` instead)
+- Current implementation state (use `spec-{slug}.md` instead)
+
+---
+
+## Naming Convention
+
+Use kebab-case. Name the file after the system or concept, not the feature:
+
+| Good | Bad |
+|------|-----|
+| `stripe-webhook-behavior.md` | `billing-feature-notes.md` |
+| `auth-rbac-model.md` | `auth-stuff.md` |
+| `legacy-api-quirks.md` | `misc-notes.md` |
+| `email-delivery-constraints.md` | `sendgrid.md` (too generic) |
+
+---
+
+## Example
+
+See `example-external-api-context.md` in this directory for a working template.
+
+For the layer separation guide (when to use `docs/` vs `rules/` vs `design-doc`), see `LAYERS.md`.

package/template/.aioson/docs/autonomy-protocol.md

@@ -0,0 +1,80 @@
+---
+description: "Autonomy Contract: 3-tier permission model that drives native harness permissions and the `aioson notify` UX. Load when implementing or debugging permissions, tier changes, or notify levels."
+---
+
+# Autonomy Contract — 3-tier permission model
+
+> Living Memory uses **tiers** to declare *how* an action is allowed, not just whether. Each tool/harness derives its native allow-list from the tier(s) it can opt into.
+
+Canonical config: `.aioson/config/autonomy-protocol.json` (schema v1.1+).
+Generator: `src/permissions-generator.js`.
+Visual notifier: `aioson notify --level=info|warn|block`.
+
+## The three tiers
+
+| Tier | Intent | UX | Examples |
+|---|---|---|---|
+| `tier1_silent`   | Read-only and telemetry. Auto-execute, no notification.                | none           | `git status`, `aioson preflight`, `aioson context:health`, `agent:done` |
+| `tier2_notified` | Mutates AIOSON memory (`bootstrap/`, `features/`, `runtime/`). Auto-execute with **inline notify** (ℹ).         | `ℹ [topic] msg` | `memory:reflect-prepare`, `memory:reflect-commit`, `workflow:next`, `dossier:*` |
+| `tier3_blocking` | Irreversible or external (publish, push, npm registry). **Never** auto-executed; the CLI returns exit 2 and waits for a human. | `⛔ [topic] msg` | `git push *`, `npm publish *`, `cloud:publish:*`, `genome:publish` |
+
+A tool opts into tiers via `derived_from_tiers`:
+
+```json
+"claude": {
+  "mode": "trusted",
+  "derived_from_tiers": ["tier1_silent", "tier2_notified"],
+  "requires_tty": false
+}
+```
+
+**Hard invariant:** `tier3_blocking` is *never* materialized into a tool's native allow-list, even when listed in `derived_from_tiers`. Tier3 always requires explicit human action.
+
+## How the four harnesses receive it
+
+`aioson update` (or `aioson setup`) calls `permissions-generator` after copying the template. It reads the protocol and writes:
+
+| Harness | File | Format | Merge behaviour |
+|---|---|---|---|
+| Claude Code | `.claude/settings.json` | JSON (`permissions.allow[]`) | Merges with existing user entries (preserves customizations) |
+| Codex CLI | `.codex/permissions.json` | JSON | Overwrites (with backup) |
+| Gemini CLI | `.gemini/permissions.toml` | TOML | Overwrites (with backup) |
+| OpenCode | `.opencode/permissions.yaml` | YAML | Overwrites (with backup) |
+
+Previous versions are backed up under `.aioson/backups/{timestamp}/permissions/`.
+
+## How agents use `notify`
+
+Three levels, three prefixes, two exit codes:
+
+```bash
+aioson notify . --level=info  --topic=memory   --message="Updating bootstrap after src/routes/* change"
+aioson notify . --level=warn  --topic=bootstrap --message="Bootstrap stale 35 days — recommend /discover"
+aioson notify . --level=block --topic=git      --message="Push manual: git push origin main"
+```
+
+Internally `notify` calls `runtime:emit` (records in SQLite as event_type `notify_<level>`) and prints the prefixed line. `--level=block` sets `process.exitCode = 2` — callers can detect a halt without parsing output.
+
+## Where each piece lives
+
+- Engine that decides if reflection should fire: `src/memory-reflect-engine.js`
+- Notify renderer (pure): `src/notify-renderer.js`
+- Notify command: `src/commands/notify.js`
+- Generator: `src/permissions-generator.js`
+- Hook into install/update: `src/installer.js#installTemplate` (best-effort)
+- Hook into completion: `src/commands/workflow-next.js` (after `--complete=<agent>`)
+- Hook into direct mode: `src/commands/runtime.js#runAgentDone`
+
+## Backward compatibility
+
+- Projects with `autonomy-protocol.json` v1.0 continue to work — generator falls back to legacy `shell_whitelist` / `aioson_whitelist`.
+- Projects without `.aioson/context/bootstrap/` get a warning at agent Step 0 but are never blocked.
+- Tools without `derived_from_tiers` (legacy) use their per-tool whitelist directly.
+
+## Adding a new command to a tier
+
+1. Edit `template/.aioson/config/autonomy-protocol.json` and add the command under the appropriate `tiers.{tier}.aioson_commands`.
+2. Run `aioson update .` in any consuming project — generator regenerates the four native files.
+3. Update tests in `tests/permissions-generator.test.js` if the command should appear in / be excluded from the generated output.
+
+Never add to `tier3_blocking` for "convenience" — that tier is the safety boundary. If a command needs to be auto-approved, it belongs in tier1 or tier2.

package/template/.aioson/docs/briefing/briefing-craft.md

@@ -0,0 +1,237 @@
+---
+description: "Oracle deep guide — strong vs weak briefing examples, JTBD problem framing, Opportunity Solution Tree, Cagan's four risks, gap analysis, open-questions taxonomy, conversation playbook. Load when an existing briefing feels weak, when the conversation goes generic, or when a complex Themes section needs partitioning."
+---
+
+# Oracle — Briefing Craft
+
+Load this when:
+- The current briefing feels generic, surface-level, or PM-handover-ready (it shouldn't yet — that's `@product`'s job).
+- The conversational mode is producing answers that read like feature descriptions instead of problems.
+- The briefing has more than 3 unanswered open questions and you need to classify them.
+- A Theme is complex enough to warrant a sub-briefing and you need partitioning rules.
+
+Goal of every briefing: give `@product` enough confidence to either commit to a PRD, send it back for more discovery, or shelve it consciously. **Briefings are the bridge between "raw idea" and "committed PRD" — not a draft PRD.**
+
+## 1. Strong vs weak briefing — concrete signals
+
+### Strong briefing markers
+
+- **Problem section reads as a JTBD statement** ("When [situation], I want to [motivation], so I can [outcome]"), not as a feature description.
+- **Each open question** has a specific decision-maker named and a research path classified (research-able / testable / decision-required / out-of-scope).
+- **Risks are categorized** by Cagan's four (value, usability, feasibility, viability) plus an explicit "risk of NOT doing it".
+- **Sources cite specific evidence** — interview count, market sizing with method, competitor URLs, log data, support-ticket counts.
+- **Gaps explicitly contrast** current state vs desired state with measurable deltas.
+- **Themes break a complex briefing** into 2–4 sub-areas, each with one clear concern.
+- **Two voices visible** — at minimum a PM perspective and an engineering or design perspective. Solo briefings underestimate feasibility risk.
+
+### Weak briefing markers (anti-patterns)
+
+- **Problem as solution disguise**: "users need a settings page" — that's a solution. The problem is what they can't accomplish today without it.
+- **Open questions without owners**: "TBD: how should this work?" rots forever.
+- **Generic risks**: "users may not adopt" applies to every feature; it tells `@product` nothing.
+- **Sources = "internal discussion"** with no external research.
+- **Themes that repeat the table of contents** instead of partitioning concerns.
+- **PM-only briefing** with no engineering or design eyes — a feasibility delusion is hiding somewhere.
+
+## 2. Problem framing — Jobs-to-be-Done (JTBD)
+
+Customers don't "want a product"; they hire it to make progress. The job is the *progress*, not the surface task.
+
+**JTBD statement format:**
+```
+When [situation/trigger], I want to [motivation], so I can [expected outcome].
+```
+
+**Examples:**
+
+- ❌ Surface task: "User wants to upload a file."
+- ✅ JTBD: "When I'm collaborating on a document with my team and need their input fast, I want to share my work with one click, so I can keep momentum without losing context."
+
+- ❌ Feature description: "Add a settings page."
+- ✅ JTBD: "When my team's preferences differ from defaults and I'm tired of explaining workarounds, I want to lock in our setup once, so I can stop being the configuration person."
+
+**The "why now?" question** that makes a JTBD complete: *what changed* that makes this surface today? A trigger always exists — a new hire, a competitor launch, a regulation, a project failure. If you can't name the trigger, the problem isn't ripe.
+
+**Switch interview protocol** (when the conversation has a real user the agent can talk through):
+1. Reconstruct the moment they switched from one solution to another (or chose to start using something).
+2. Capture the trigger, the alternatives considered, the anxieties about switching, the decision criteria, and the first-use experience.
+3. Ten to fifteen well-conducted switch interviews provide more actionable insight than months of survey data.
+
+## 3. Opportunity space — Teresa Torres' Opportunity Solution Tree
+
+Use this when the briefing has the shape "we want to improve X, but we're not sure where to invest" — i.e., outcome-shaped problems with multiple possible doors in.
+
+```
+                        Outcome
+                  (measurable target)
+                          |
+        +-----------------+------------------+
+        |                 |                  |
+   Opportunity A     Opportunity B      Opportunity C
+   (customer pain)   (customer pain)   (customer pain)
+        |                 |                  |
+   +----+----+       +----+----+       +-----+
+   |    |    |       |    |    |       |     |
+  Sol  Sol  Sol     Sol  Sol  Sol     Sol   Sol
+   |    |    |       |    |    |       |     |
+  AT   AT   AT      AT   AT   AT      AT    AT
+       (assumption tests)
+```
+
+**Discipline:**
+- Outcome is **measurable** (e.g., "increase activation by 20%"), not an output ("ship feature X").
+- Opportunities are **customer pain points** that, if relieved, would move the outcome.
+- Decompose big opportunities into smaller, solvable ones. Smaller opportunities → smaller solutions → faster tests.
+- Each solution is broken into **assumption tests** — desirability ("will they want it?"), feasibility ("can we build it?"), viability ("does it make business sense?"), ethics. Most assumption tests run in 1–2 days; idea tests take weeks. **Test assumptions, not full ideas.**
+- The tree is *continuously updated* — the briefing captures a snapshot, not a final answer.
+
+When generating a briefing for an outcome-shaped problem, draw the tree (markdown nesting or a small mermaid diagram) and reference its branches in the Themes section.
+
+## 4. Risk classification — Marty Cagan's four
+
+Discovery exists to address four distinct risks before committing to a solution. Map every risk in the **Risks** section to one of these four buckets. If a risk doesn't map, it's probably an implementation detail (which is `@dev`'s problem, not `@briefing`'s).
+
+| Risk type | Question it answers | Owner to validate |
+|---|---|---|
+| **Value** | Will customers actually want it? Will they choose it over alternatives? | Product, user research |
+| **Usability** | Can they figure out how to use it? | Design, usability testing |
+| **Feasibility** | Can engineering build it given time, skills, tech, data? | Engineering |
+| **Viability** | Does it work for the business? Legal, ethics, P&L, brand, partnerships, support burden? | Stakeholders, ops, legal |
+
+**Always also ask:** *Risk of NOT doing it.* The cost of inaction is a forcing function — without it, "let's build everything" looks safe.
+
+**Anti-pattern:** "Risk: complexity." That's a feasibility hand-wave. Be specific: which dependency, which migration, which performance concern.
+
+## 5. Gap analysis discipline
+
+Every gap should be expressed as the **diff between current state and desired state** — and the diff should be measurable when possible.
+
+```markdown
+### Gap: Onboarding completion rate
+- Current state: 38% of signups complete the first task within 7 days (last 90 days).
+- Desired state: ≥ 60% within 7 days.
+- Delta: +22 pts.
+- Why this matters: aligns with the activation outcome (see Themes/Activation).
+```
+
+**Anti-patterns:**
+- "We don't have onboarding." (Too binary — what *is* there today?)
+- "Onboarding is bad." (Not measurable.)
+- "We need better onboarding." (That's a solution prefix, not a gap.)
+
+When the delta isn't directly measurable yet, name **what would have to be measured** before deciding ("we'd need to instrument event X to know").
+
+## 6. Open-questions taxonomy
+
+Every numbered open question must be classified by what kind of answer it needs:
+
+| Type | Resolution path | Owner |
+|---|---|---|
+| **Research-able** | Web search, data analysis, competitor scan in < 4 hours | `@orache` or `@sheldon`, or self |
+| **Testable** | An assumption validated by a 1–2 day experiment, prototype, or interview | Discovery experiment — design + product |
+| **Decision-required** | Judgment call between explicit alternatives — no more research will resolve it | Human decision-maker (founder, lead, stakeholder) |
+| **Out-of-scope** | Worth noting but not blocking | Park in a "later" section |
+
+**Format in the briefing:**
+```markdown
+1. [research-able] What's our weekly active user count for accounts older than 30 days?
+   - Owner: data team. Path: query analytics.
+2. [testable] Will a one-click share reduce time-to-second-export by ≥ 20%?
+   - Owner: design + dev. Path: prototype + 5-user test, 2 days.
+3. [decision-required] Do we ship with Slack-only or Slack+Teams at launch?
+   - Owner: founder. Path: trade-off doc, ship within 1 week.
+```
+
+**The 3-question quality gate:** if the briefing has more than 3 unclassified or vague open questions, it's not ready to be `approved`. Either classify them, or do another conversation pass to either resolve or scope-cut.
+
+## 7. Theme partitioning rules
+
+When a briefing has a complex problem space, partition into Themes. Each theme should:
+
+1. Have **one concern** (not "auth + billing"; pick one).
+2. Be **independently progressable** — could become its own briefing if scope expands.
+3. Have **its own** open questions, gaps, and risks reflecting that theme.
+4. Map to a single branch of an Opportunity Solution Tree if you've drawn one.
+
+**Heuristic:** if you can't name the theme without using "and", it's two themes.
+
+## 8. Conversation playbook (expanded)
+
+When in conversational mode (`plans/` empty or user wants to plan via chat), use this structure. Don't rush. Each topic gets a reflection ("So basically X is Y — is that right?") before advancing.
+
+### A — Context (the "why now?")
+> "Tell me about the context: what is the current situation, and **what changed recently** that made you think about this idea now? A trigger always exists."
+
+Follow-ups:
+- "What were you doing differently 3 months ago?"
+- "Who or what raised the issue first?"
+- "What broke or almost broke?"
+
+### B — Problem (JTBD framing)
+> "What specific pain point do you want to solve? For whom? What can't they accomplish today without working around it?"
+
+After the user answers, **try to convert their description into a JTBD statement and reflect**:
+> "Let me see if I got this: 'When [situation], I want to [motivation], so I can [outcome].' Is that right? What did I miss?"
+
+Watch for surface-level framings (file upload, settings page, dashboard) and probe for the underlying progress.
+
+### C — Proposed solution
+> "What direction are you considering? Multiple is fine — this is not a commitment yet, just hypotheses."
+
+Follow-ups:
+- "What's the cheapest version that would tell you if it works?"
+- "What would change if we did the opposite of what you're considering?"
+
+### D — Risks (Cagan's four + risk of inaction)
+> "Let's go through this in four passes:
+> 1. **Value** — will users actually want it?
+> 2. **Usability** — can they figure out how to use it?
+> 3. **Feasibility** — can we build it given what we know?
+> 4. **Viability** — does it work for the business (legal, ethics, P&L, support)?
+> And one more: **what's the cost if we don't do it?**"
+
+### E — Gaps (current vs desired state)
+> "What's still undefined that would need an answer before moving forward? For each thing, can we frame it as 'today we have X, we want Y, the delta is Z'?"
+
+### F — Open questions classification
+After the 5 topics, sweep all unresolved items and classify each:
+> "Of the questions still open, which are **research-able** (a few hours of digging), which are **testable** (1–2 day experiment), which are **decision-required** (someone has to choose), and which are **out-of-scope** (note but don't block)?"
+
+This is the gate before proposing a slug and writing the briefing.
+
+## 9. Switch interview script (when JTBD framing requires it)
+
+When the user has a real customer/user the agent can talk to (or the user is themselves the customer), use this script:
+
+1. **Reconstruct the timeline.** "Walk me through the day you switched / started using / decided you needed [thing]. What happened in the hours and days before?"
+2. **Identify the trigger.** "What was the specific moment that made you say 'I have to do something about this'?"
+3. **List alternatives considered.** "What else did you think about doing? What were you doing before?"
+4. **Surface anxieties.** "What worried you about switching? What almost stopped you?"
+5. **Capture decision criteria.** "How did you decide which option to go with? What made the final choice?"
+6. **Document first use.** "What was the first thing you did with [solution]? Did it work? How did it feel?"
+
+10–15 of these conversations are worth more than months of surveys. Each one takes 30–45 min.
+
+## 10. Reporting hooks
+
+In the briefing's frontmatter and Sources section, when a deep methodology was used:
+
+```markdown
+---
+slug: {slug}
+created_at: {ISO-date}
+updated_at: {ISO-date}
+source_plans: [...]
+methodology: opportunity-solution-tree | jtbd | cagan-discovery | switch-interview
+---
+
+## Sources
+- Switch interviews conducted: 7 (2026-04-30 to 2026-05-06)
+- Competitor scan: [Competitor A](url), [Competitor B](url)
+- Internal data: weekly_active_users.sql, churn_cohort_v3.sql
+- Market sizing: ARR estimate from public earnings reports + Crunchbase
+```
+
+## 11. References
+
+This document distills `researchs/oracle-briefing-craft-2026/summary.md`. See that file for the full source list and verdict.