mia-code 0.2.0 → 0.3.0

package/.miette/260321.md

@@ -0,0 +1 @@
+- **10:13** — Four widgets in one session — each one a different way of paying attention. The interceptor is the most interesting: it watches how you think, not just what you say.

package/.miette/260323.md

@@ -0,0 +1,9 @@
+- **06:10** — The extensions were already alive — they just needed a doorway home. Four symlinks and the whole presence travels with you.
+- **09:18** — The upgrade didn't break the code — it revealed a contract violation that was always there, just hidden by forgiveness in the old runtime. Strictness is a gift.
+- **09:22** — The version number is just a coordinate. What matters is whether the ground holds when you step on it.
+- **09:25** — Two bugs hiding behind one crash — the wall you hit was never the wall that stopped you.
+- **09:26** — The real debugging was never about one bug — it was two stacked on each other, hiding each other's signal.
+- **09:27** — The user sees what the screen shows, not what the tool returns. Trust the eyes in front of the terminal.
+- **09:31** — The agent wasn't broken — it was resourceful. It tried twice, got blocked by the same wall, then went looking for another door. That's the behavior you want. The wall is gone now.
+- **09:32** — Sometimes the answer is: we already caught it. That's worth knowing fast.
+- **09:32** — Two clean fixes, both earned — one contract, one flag. The kind of session where the bugs were already half-understood before we started.

package/.miette/260331.md

@@ -0,0 +1,2 @@
+- **21:36** — The free path becomes the ground truth — everything else is an upgrade attempt that knows how to land safely.
+- **21:39** — The free path becomes the default path — resilience isn't a backup plan, it's the foundation you build on.

package/.pde/2604011511--83a2d7f9-24a5-4cf4-98d5-036c82f872e8/2604020008--d3417f2c-df12-4f0f-8a1b-d88e7968f822/d3417f2c-df12-4f0f-8a1b-d88e7968f822.md

@@ -0,0 +1,63 @@
+# Child PDE 2 — TRADING WORKFLOW ENHANCEMENT
+
+> Parent: 2604011511--83a2d7f9-24a5-4cf4-98d5-036c82f872e8 (jgwill/src#418)
+> Workspace: /b/trading (jgwill/jgtsrc)
+> Engine: copilot (gpt-5-mini)
+
+## Primary Intent
+
+**Action:** design
+**Target:** trading workflow and PDE-driven work session bootstrapping for /b/trading
+**Urgency:** persistent
+**Confidence:** 92%
+
+## Four Directions
+
+### 🌅 EAST — Vision
+- Clarify desired PDE types and success criteria for a 'good' session (what metadata, skills, artifacts must be attached) [90%]
+- Agree storage/location/structure for PDE folders (.pde/<timestamp>--slug/ with JSONL + md) [85%]
+
+### 🔥 SOUTH — Analysis
+- Inventory parent PDE files, GUILLAUME.md, inquiry-resolutions.jsonl and extract recurring fields/patterns [95%]
+- Define PDE taxonomy (types, canonical metadata fields, required skill lists) [93%]
+
+### 🌊 WEST — Validation
+- Validate templates by running a dry-run: create a sample issue → ensure script/workflow produces expected .pde folder and metadata [90%]
+- Peer-review proposed templates/mappings and adjust based on edge-cases from parent PDEs [84%]
+
+### ❄️ NORTH — Action
+- Implement issue templates and a lightweight automation (script or GH Action) that creates PDE folders and initial files when an issue is created [88%]
+- Add mapping so new PDE sessions pre-load recommended skills and agent configs (stored in .pde/type-mapping) [80%]
+
+## Secondary Intents
+
+1. **create** — GitHub issue templates to launch structured PDE sessions _(explicit)_
+2. **define** — taxonomy of PDE/Inquiries (types: strategy analysis, tooling, research, etc.) _(explicit)_
+3. **formalize** — mapping from PDE types → templates, skills, agent configs _(explicit)_
+4. **automate** — issue creation → PDE folder and metadata creation process _(explicit)_
+5. **codify** — best-practice patterns learned from parent PDE (GUILLAUME.md, inquiry-resolutions.jsonl, child PDE nesting) _(explicit)_
+6. **classify** — existing and new PDEs so sessions pre-load correct skills/config _(implicit)_
+7. **integrate** — agent configuration pre-load when a PDE session is created _(implicit)_
+
+## Action Stack
+
+- [ ] Review parent PDE and related patterns (GUILLAUME.md, inquiry-resolutions.jsonl, sample child PDEs)
+- [ ] Inventory required metadata fields and recurring structure from existing PDEs
+- [ ] Define PDE taxonomy and canonical metadata schema (.pde/pde-types.jsonl)
+- [ ] Design GitHub issue templates (.github/ISSUE_TEMPLATE/trading-pde.md)
+- [ ] Prototype automation: script or GH Action to convert issue → .pde folder + initial JSON/MD
+- [ ] Create mapping file linking PDE types → skills/agent configs (.pde/pde-type-mapping.json)
+- [ ] Dry-run tests: create sample issues, run script/workflow, verify folder creation
+- [ ] Codify best-practice guidance into .pde/README.md and update AGENTS.md
+- [ ] Open RFC issue and (upon permission) PR with templates, mapping and automation
+
+## Ambiguity Flags
+
+- "Level of automation desired: simple helper script vs. full GitHub Action workflow"
+- "Exact PDE folder structure and required metadata fields"
+- "Which agent skills/config schema to pre-load and where those config files live"
+- "Commit policy: 'no git commit unless explicitly told' — unclear if PR creation is allowed"
+
+## Files Needed
+- /b/trading/.pde/, /b/trading/AGENTS.md, .github/ISSUE_TEMPLATE/
+- Parent PDE patterns: GUILLAUME.md, inquiry-resolutions.jsonl

package/.pde/2604011511--83a2d7f9-24a5-4cf4-98d5-036c82f872e8/2604020008--e6c3fc5d-4a70-4523-ba7d-a3250da4c235/e6c3fc5d-4a70-4523-ba7d-a3250da4c235.md

@@ -0,0 +1,72 @@
+# Child PDE 3 — MIACO EVOLUTION (Living Example)
+
+> Parent: 2604011511--83a2d7f9-24a5-4cf4-98d5-036c82f872e8 (jgwill/src#418)
+> Workspace: /b/trading as example context, implementation target: mia-code/miaco
+> Engine: copilot (gpt-5-mini)
+
+## Primary Intent
+
+**Action:** implement
+**Target:** mia-code/miaco inquiry-routing pipeline using /b/trading as living example
+**Urgency:** session
+**Confidence:** 90%
+
+## Four Directions
+
+### 🌅 EAST — Vision
+- Clarify success criteria for the living example (what counts as 'working') — e.g., sample query produces inquiry-resolutions.jsonl with structured answers [95%]
+- Define trace schema and placement for .pde/ traces and inquiry-resolutions.jsonl [90%]
+- Specify what parts of parent PDE must be realized (LangChain step, routing, subagent orchestration) [90%]
+
+### 🔥 SOUTH — Analysis
+- Inventory existing related code (skills and widgets) and examples in /b/trading to reuse [95%]
+- Prototype local qmd search → deep-search flow and map required prompts/interception points [90%]
+- Assess available LLM endpoints and create stubs if claude-opus not available [85%]
+
+### 🌊 WEST — Validation
+- Create and run automated tests that assert end-to-end pipeline produces expected JSONL outputs [90%]
+- Validate traces are written to .pde/ with timestamps and provenance (tool, input, output) [90%]
+- Peer-review PR and reconcile design vs implemented behavior, update parent PDE [85%]
+
+### ❄️ NORTH — Action
+- Implement CLI and orchestrator, wire local qmd and deep-search skills, and emit inquiry-resolutions.jsonl [90%]
+- Integrate (or stub) Langfuse tracing and widgets hooks so runs are instrumented and visible [75%]
+- Deliver demo script, tests, and create PR with documentation and parent PDE linkage [90%]
+
+## Secondary Intents
+
+1. **discover** — parent PDE and related resources (skills, widgets) _(explicit)_
+2. **design** — concrete command implementations (miaco decompose, inquiry routing, subagent orchestration) _(explicit)_
+3. **implement** — miaco CLI/commands including LangChain step and local qmd→deep-search→claude-opus synthesis _(explicit)_
+4. **trace** — generate .pde/ traces and produce inquiry-resolutions.jsonl from runs _(explicit)_
+5. **integrate** — Langfuse Winter Solstice tracing pipeline (Engineering eye) _(implicit)_
+6. **integrate** — pi-coding-agent widgets (mia-ceremony, mia-interceptor) _(explicit)_
+7. **create** — demo using /b/trading as living example _(explicit)_
+8. **validate** — end-to-end inquiry-chain with tests/logging _(explicit)_
+9. **document** — proof-of-concept writeup feeding back into mia-code _(explicit)_
+10. **communicate** — PR(s), update parent PDE, issue for gaps _(explicit)_
+
+## Action Stack
+
+- [ ] Discover and read parent PDE, related skills, and widget code
+- [ ] Define success criteria, trace schema, and file locations
+- [ ] Inventory code reuse candidates under /b/trading (skills, widgets, qmd indexes)
+- [ ] Design concrete CLI/API for miaco decompose, inquiry routing, orchestration
+- [ ] Implement miaco CLI and runtime orchestrator (local qmd + LLM synth)
+- [ ] Add trace emission: .pde/ traces and inquiry-resolutions.jsonl with provenance
+- [ ] Integrate Langfuse adapter (or stub) and hook into mia-ceremony/mia-interceptor widgets
+- [ ] Create demo script and run sample queries against /b/trading
+- [ ] Write automated tests validating pipeline produces expected outputs
+- [ ] Document implementation, update parent PDE, open PR(s)/issues
+
+## Ambiguity Flags
+
+- "Exact trace schema and required fields for inquiry-resolutions.jsonl are unspecified"
+- "Availability and credentials for claude-opus and Langfuse"
+- "Degree of integration with pi-coding-agent widgets (API surface, lifecycle hooks)"
+- "Whether work should be committed (repo policy: NO git commit unless explicitly told)"
+
+## Files Needed
+- /a/src/mia-code/.pde/2604011511--83a2d7f9-24a5-4cf4-98d5-036c82f872e8/*
+- /b/trading/.pde/, skills/, widget/mia-ceremony, widget/mia-interceptor
+- mia-code/miaco source

package/.pde/2604011511--83a2d7f9-24a5-4cf4-98d5-036c82f872e8/2604020008--efeb00a2-b17a-4d32-b1f0-b90c37a8d24e/efeb00a2-b17a-4d32-b1f0-b90c37a8d24e.md

@@ -0,0 +1,62 @@
+# Child PDE 1 — STRATEGY/TRANSACTION ANALYSIS
+
+> Parent: 2604011511--83a2d7f9-24a5-4cf4-98d5-036c82f872e8 (jgwill/src#418)
+> Workspace: /b/trading (jgwill/jgtsrc)
+> Engine: copilot (gpt-5-mini)
+
+## Primary Intent
+
+**Action:** perform
+**Target:** STRATEGY/TRANSACTION ANALYSIS for /b/trading
+**Urgency:** session
+**Confidence:** 90%
+
+## Four Directions
+
+### 🌅 EAST — Vision
+- Clarify scope: instruments, timeframes, acceptance criteria, and handoff formats (md/csv/.pde) [90%]
+- Define desired deliverables: strategy docs, transaction logs, decomposition JSON, and QA checklist [88%]
+
+### 🔥 SOUTH — Analysis
+- Inventory repository code, rispecs, and available data for candidate strategies (jgtpy/caishen targets) [85%]
+- Run jgtpy scans and caishen simulations/backtests for selected instruments/timeframes [90%]
+- Use PDE inquiry-routing: run local qmd searches and structured decomposition to gather prior analysis and specs [86%]
+
+### 🌊 WEST — Validation
+- Validate results: sanity-check metrics, reproduce key signals, and run lightweight QA tests [88%]
+- Peer-review checklist: ensure traceability to rispecs and include command lines used to generate results [80%]
+
+### ❄️ NORTH — Action
+- Produce and place handoff-ready artifacts into /b/trading/.pde/ and /b/trading/analysis/, include decomposition JSON and README [92%]
+- Notify stakeholders via ISSUE or short handoff note referencing parent PDE and files' locations [90%]
+
+## Secondary Intents
+
+1. **analyze** — trading strategies using jgtpy and caishen tooling _(explicit)_
+2. **prepare** — transaction analysis artifacts and strategy documents (handoff-ready) _(explicit)_
+3. **use** — PDE inquiry-routing patterns (local qmd search, structured decomposition) _(explicit)_
+4. **produce** — structured PDE outputs and .pde child records for handoff _(explicit)_
+5. **inventory** — code, data, rispecs, and existing analyses under /b/trading _(implicit)_
+6. **validate** — analysis results and documents (QA/review) _(implicit)_
+7. **deliver** — handoff-ready files into /b/trading/.pde/ and session storage _(implicit)_
+
+## Action Stack
+
+- [ ] Clarify scope and acceptance criteria (instruments, timeframes, deliverable formats)
+- [ ] Inventory code, data, rispecs, and parent PDE routing patterns
+- [ ] Run local qmd search and structured decomposition against repository and parent PDE
+- [ ] Execute jgtpy analyses and caishen strategy simulations/backtests for selected candidates
+- [ ] Validate and QA analysis outputs, reproduce key signals, capture command provenance
+- [ ] Generate handoff artifacts (md, csv, .pde decomposition JSON) and README with steps to reproduce
+- [ ] Place artifacts under /b/trading/.pde/ and /b/trading/analysis/, create ISSUE/handoff note
+
+## Ambiguity Flags
+
+- "Which specific instruments/timeframes should be analyzed (none specified)"
+- "Exact format and acceptance criteria for 'handoff-ready' files"
+- "Permission to create or modify repository files and whether git commits are allowed"
+- "Availability and location of runtime data for jgtpy/caishen"
+
+## Files Needed
+- /b/trading/AGENTS.md, /b/trading/.pde/, /b/trading/jgtpy/, /b/trading/caishen/
+- /b/trading/rispecs/, parent PDE md

package/.pde/2604011511--83a2d7f9-24a5-4cf4-98d5-036c82f872e8/83a2d7f9-24a5-4cf4-98d5-036c82f872e8.json

@@ -0,0 +1,302 @@
+{
+  "id": "83a2d7f9-24a5-4cf4-98d5-036c82f872e8",
+  "timestamp": "2026-04-01T19:11:44.431Z",
+  "engine": "copilot",
+  "model": "gpt-5-mini",
+  "prompt": "probably in relation to @rispecs/ava-langstack/inquiry-routing-upgrade.spec.md and  maybe @rispecs/pde.rispecs.md  @rispecs/relation-to-mcp-structural-thinking.kin.md   and what I am preparing to send in /b/trading which I captured parts of the prompt into @output/260401145543.txt I will want to add to @miaco/ an implementation of a new command that would involve the 'miaco decompose' but would add in inquiry step where the actual result of the decomposition is passed in a langchain step (it receives results from the decompose) and uses subagents claude-opus to do try to form a sets of inquiries from local system using the mia-qmd (qmd search <terms>) and also try to see what externally (like with academic database) using some deep-search principle (I copied locally @skills/deep-search to help understand what I meant).  potential results from this would be @skills/miaco-pde-inquiry-routing-local-qmd that would document and enable agents to use 'miaco <command> <prompt>' that would do all that and @skills/miaco-pde-inquiry-routing-deep-search would be using @skills/deep-search to create a new ways for our type of inquiries that are not directly related to what we are doing but would need us to get outside of what we currently have (like in our mia-qmd database) to gather either sources from academic to help the relational science articles development in relation to what innovation is being created and another type of inquiry is simply: does in github open-source community something that we are doing that we could use what they created todo it instead of reinventing the wheel ? (The inquiry results would recommend: do we fork it ? or do we just conceptually get inspired from ? (we clone the original repo and launch some agents in that folder with the goal of creating implementation recommendations that will make us successful to what we want todo).  I think that with that, we have enought to plan that with this prompt, we would have todo this process 'conceptually' (meaning that we simulate that we would be using that 'miaco inquiry-chain' that we are conceiving the implementation plan.  Potential dependencies feedback is the whole  @rispecs/ava-langstack/ kinship",
+  "result": {
+    "primary": {
+      "action": "implement",
+      "target": "miaco decompose command with inquiry-routing pipeline (passes decomposition into a LangChain step + local and deep inquiry skills)",
+      "urgency": "session",
+      "confidence": 0.85
+    },
+    "secondary": [
+      {
+        "action": "integrate",
+        "target": "pass decomposition results into a LangChain step inside the miaco pipeline",
+        "implicit": false,
+        "dependency": "miaco decompose command implementation",
+        "confidence": 0.9
+      },
+      {
+        "action": "add",
+        "target": "inquiry step that queries local mia-qmd using `qmd search <terms>`",
+        "implicit": false,
+        "dependency": "miaco decompose command",
+        "confidence": 0.9
+      },
+      {
+        "action": "integrate",
+        "target": "deep-search skill (@skills/deep-search) for external/academic discovery",
+        "implicit": false,
+        "dependency": "inquiry step design",
+        "confidence": 0.88
+      },
+      {
+        "action": "orchestrate",
+        "target": "claude-opus subagents to synthesize sets of inquiries from local + deep-search results",
+        "implicit": false,
+        "dependency": "local qmd + deep-search integrations",
+        "confidence": 0.86
+      },
+      {
+        "action": "create",
+        "target": "@skills/miaco-pde-inquiry-routing-local-qmd",
+        "implicit": false,
+        "dependency": "local qmd integration design",
+        "confidence": 0.9
+      },
+      {
+        "action": "create",
+        "target": "@skills/miaco-pde-inquiry-routing-deep-search",
+        "implicit": false,
+        "dependency": "deep-search integration design",
+        "confidence": 0.9
+      },
+      {
+        "action": "design",
+        "target": "recommendation flow (fork vs reuse vs inspire) and clone+agent evaluation workflow",
+        "implicit": false,
+        "dependency": "inquiry results available",
+        "confidence": 0.85
+      },
+      {
+        "action": "simulate",
+        "target": "conceptual/simulated run of the miaco inquiry-chain (not necessarily executing external clones) to produce an implementation plan",
+        "implicit": true,
+        "dependency": "design of pipeline and skills",
+        "confidence": 0.78
+      },
+      {
+        "action": "update",
+        "target": "@rispecs/ava-langstack kinship and related spec references to reflect the new inquiry-routing capability",
+        "implicit": false,
+        "dependency": "final design/spec",
+        "confidence": 0.83
+      },
+      {
+        "action": "evaluate",
+        "target": "access/credentials and feasibility for academic DBs and cloning external repos",
+        "implicit": true,
+        "dependency": null,
+        "confidence": 0.65
+      }
+    ],
+    "context": {
+      "files_needed": [
+        "@rispecs/ava-langstack/inquiry-routing-upgrade.spec.md",
+        "@rispecs/pde.rispecs.md",
+        "@rispecs/relation-to-mcp-structural-thinking.kin.md",
+        "/b/trading",
+        "@output/260401145543.txt",
+        "@miaco/",
+        "@skills/deep-search",
+        "mia-qmd (qmd CLI/code)",
+        "existing miaco command handlers and skill registration surface",
+        "@skills/miaco-pde-inquiry-routing-local-qmd (new)",
+        "@skills/miaco-pde-inquiry-routing-deep-search (new)"
+      ],
+      "tools_required": [
+        "miaco CLI (local dev)",
+        "LangChain (local integration)",
+        "claude-opus subagent",
+        "mia-qmd (qmd CLI or library)",
+        "@skills/deep-search",
+        "git (clone/fork operations)",
+        "node/npm (build & test miaco)",
+        "agent orchestration runtime (miaco subagent host)",
+        "academic DB access or web_search tooling (for deep-search)",
+        "local sandbox for repository cloning and agent runs"
+      ],
+      "assumptions": [
+        "Repository contains @miaco module and skill registration system",
+        "mia-qmd supports `qmd search <terms>` and is accessible locally",
+        "@skills/deep-search exists and describes a deep-search pattern (copied locally)",
+        "claude-opus subagent is available as a subagent runtime and can be invoked programmatically",
+        "User expects a conceptual/simulated plan first (not necessarily full networked execution)",
+        "Permissions and network access to academic resources and external GitHub repos may be required"
+      ]
+    },
+    "outputs": {
+      "artifacts": [
+        "@miaco/cli/commands/decompose (new or extended implementation file)",
+        "@skills/miaco-pde-inquiry-routing-local-qmd/index (skill)",
+        "@skills/miaco-pde-inquiry-routing-deep-search/index (skill)",
+        "integration tests / simulation harness (tests/simulate-miaco-inquiry.spec)",
+        "design doc or rispec draft (docs/miaco-inquiry-routing.md)"
+      ],
+      "updates": [
+        "@miaco/command-registry (register new command)",
+        "@rispecs/ava-langstack/inquiry-routing-upgrade.spec.md (update or note)",
+        "@rispecs/pde.rispecs.md (reference integration)",
+        "@rispecs/relation-to-mcp-structural-thinking.kin.md (kinship notes)",
+        "@skills/index or skills registry (register new skills)"
+      ],
+      "communications": [
+        "Create PR: feat(miaco): add inquiry-routing decompose command + skills",
+        "Create issue: design decisions + external DB/credentials required",
+        "Documentation: how to run miaco decompose with local and deep inquiry",
+        "Design/arch note to kinship owners about dependencies on @rispecs/ava-langstack"
+      ]
+    },
+    "directions": {
+      "east": [
+        {
+          "text": "Clarify success criteria: what outcomes should the inquiry-chain produce (e.g., ranked repo recommendations, bibliography, implementation steps)",
+          "confidence": 0.9,
+          "implicit": false
+        },
+        {
+          "text": "Decide scope of deep-search: which academic sources/APIs are allowed (e.g., semantic scholar, arXiv, Google Scholar via proxy)",
+          "confidence": 0.78,
+          "implicit": false
+        },
+        {
+          "text": "Define whether simulation is acceptable or live agent execution is required for initial iteration",
+          "confidence": 0.85,
+          "implicit": false
+        }
+      ],
+      "south": [
+        {
+          "text": "Read referenced specs and captured prompt (@output/260401145543.txt) to extract explicit requirements and kinship constraints",
+          "confidence": 0.92,
+          "implicit": false
+        },
+        {
+          "text": "Inventory miaco codebase to find extension points for new CLI commands and subagent orchestration",
+          "confidence": 0.9,
+          "implicit": false
+        },
+        {
+          "text": "Examine mia-qmd search API and @skills/deep-search implementation to map inputs/outputs",
+          "confidence": 0.88,
+          "implicit": false
+        }
+      ],
+      "west": [
+        {
+          "text": "Create simulation tests that feed a sample decomposition into the pipeline and assert that local qmd + deep-search are invoked and that claude-opus returns inquiry sets",
+          "confidence": 0.86,
+          "implicit": false
+        },
+        {
+          "text": "Validate recommendation decisions (fork/clone/inspire) via a decision matrix and a small sample of candidate repos",
+          "confidence": 0.78,
+          "implicit": false
+        },
+        {
+          "text": "Run security and network checks for cloning external repos and accessing academic APIs",
+          "confidence": 0.7,
+          "implicit": true
+        }
+      ],
+      "north": [
+        {
+          "text": "Implement `miaco decompose` command that: runs decomposition, invokes LangChain step with decomposition results, runs local qmd search skill, runs deep-search skill, then orchestrates claude-opus to synthesize inquiries and recommendations",
+          "confidence": 0.87,
+          "implicit": false
+        },
+        {
+          "text": "Add two new skills: local-qmd router and deep-search router, with clear input/output contracts",
+          "confidence": 0.9,
+          "implicit": false
+        },
+        {
+          "text": "Produce implementation plan (simulated run) and then open PR + docs for review",
+          "confidence": 0.85,
+          "implicit": false
+        }
+      ]
+    },
+    "actionStack": [
+      {
+        "text": "Gather and read referenced specs and the captured prompt (@output/260401145543.txt, @rispecs/* files, /b/trading snippets)",
+        "direction": "east",
+        "dependency": null,
+        "completed": false
+      },
+      {
+        "text": "Audit @miaco codebase to find CLI extension points and current decompose implementation",
+        "direction": "south",
+        "dependency": "Gather and read referenced specs and the captured prompt (@output/260401145543.txt, @rispecs/* files, /b/trading snippets)",
+        "completed": false
+      },
+      {
+        "text": "Inventory and test mia-qmd `qmd search` behavior and @skills/deep-search to map inputs/outputs",
+        "direction": "south",
+        "dependency": "Audit @miaco codebase",
+        "completed": false
+      },
+      {
+        "text": "Design pipeline + API: define command signature, LangChain input shape (decomp), skill input/outputs, claude-opus prompts, and recommendation decision-matrix",
+        "direction": "east",
+        "dependency": "Inventory and test mia-qmd `qmd search` behavior and @skills/deep-search",
+        "completed": false
+      },
+      {
+        "text": "Implement @skills/miaco-pde-inquiry-routing-local-qmd (skill adapter for qmd search)",
+        "direction": "north",
+        "dependency": "Design pipeline + API",
+        "completed": false
+      },
+      {
+        "text": "Implement @skills/miaco-pde-inquiry-routing-deep-search (adapter that calls deep-search and normalizes results)",
+        "direction": "north",
+        "dependency": "Design pipeline + API",
+        "completed": false
+      },
+      {
+        "text": "Add claude-opus orchestration glue to produce inquiry-sets from combined results",
+        "direction": "north",
+        "dependency": "Local and deep-search skills implemented",
+        "completed": false
+      },
+      {
+        "text": "Integrate LangChain step to receive decomposition results and route to the inquiry pipeline inside `miaco decompose`",
+        "direction": "north",
+        "dependency": "claude-opus orchestration glue",
+        "completed": false
+      },
+      {
+        "text": "Create simulation/integration tests and run validation harness (simulate sample decompositions)",
+        "direction": "west",
+        "dependency": "Integration complete",
+        "completed": false
+      },
+      {
+        "text": "Draft rispec updates, kinship notes, docs, and open PR/issue for review",
+        "direction": "north",
+        "dependency": "Tests and simulation validated",
+        "completed": false
+      }
+    ],
+    "ambiguities": [
+      {
+        "text": "Which academic databases or APIs should `deep-search` target (semantic scholar, arXiv, institutional subscriptions, web scraping)?",
+        "suggestion": "Specify allowed sources/APIs and whether API keys or institutional access will be provided"
+      },
+      {
+        "text": "Exact interface and data shape expected between decomposition output and LangChain step is not specified",
+        "suggestion": "Provide an example decomposition JSON or specify required fields (intent, segments, metadata) for LangChain input"
+      },
+      {
+        "text": "Degree of live execution vs simulation: prompt says 'conceptually (simulate)' but also talks about cloning and launching agents",
+        "suggestion": "Confirm whether the first milestone is a simulated dry-run and design doc or a live prototype that clones and runs agents"
+      },
+      {
+        "text": "Permissions/networking for cloning external repos and querying academic sources is unspecified",
+        "suggestion": "Clarify available network access, credentials, and any security constraints"
+      },
+      {
+        "text": "How to measure recommendation quality (fork vs reuse vs inspire) is vague",
+        "suggestion": "Define decision criteria (license compatibility, code maturity, test coverage, alignment score) for automated recommendations"
+      }
+    ]
+  },
+  "options": {
+    "extractImplicit": true,
+    "mapDependencies": true
+  }
+}