@oriro/orirocli 0.1.8 → 0.1.11

package/skills/craft/become-an-ai-engineer-26/community/BUILDS.md

@@ -1,13 +1,13 @@
-# Community Builds
-
-Shipped something using this roadmap? Add it here via PR. One line each:
-
-```
-- [Project name](link) — what it does · which phase/project it proves · @yourhandle
-```
-
-See [CONTRIBUTING.md](../CONTRIBUTING.md).
-
----
-
-<!-- Add your build below this line -->
+# Community Builds
+
+Shipped something using this roadmap? Add it here via PR. One line each:
+
+```
+- [Project name](link) — what it does · which phase/project it proves · @yourhandle
+```
+
+See [CONTRIBUTING.md](../CONTRIBUTING.md).
+
+---
+
+<!-- Add your build below this line -->

package/skills/craft/become-an-ai-engineer-26/community/DISCUSSIONS.md

@@ -1,8 +1,8 @@
-# Discussions
-
-Open questions, research threads, and ongoing conversations. Use GitHub Issues
-for anything actionable; use this file to log durable findings worth keeping.
-
-## Open threads
-
-- _(none yet — start one)_
+# Discussions
+
+Open questions, research threads, and ongoing conversations. Use GitHub Issues
+for anything actionable; use this file to log durable findings worth keeping.
+
+## Open threads
+
+- _(none yet — start one)_

package/skills/craft/become-an-ai-engineer-26/phases/phase-0-mental-models/README.md

@@ -1,14 +1,14 @@
-# Phase 0 — Build Correct Mental Models (Weeks 1-2)
-
-**Do not write a single line of agent code yet.**
-
-Read `SKILL.md` → _Phase 0_. By the end you should be able to explain, in your
-own words and without looking:
-
-- Workflow vs Agent (and why building an agent when you need a workflow costs 10×)
-- The 5 workflow patterns (chaining, routing, parallelisation, orchestrator-worker, evaluator-optimiser)
-- The 4 context primitives (Write, Select, Compress, Isolate)
-- The harness stack (model = CPU, harness = OS)
-
-**Deliverable:** the 2-page concept doc. Use [`project/TEMPLATE.md`](project/TEMPLATE.md).
-If you can't write it without looking, you haven't understood it yet.
+# Phase 0 — Build Correct Mental Models (Weeks 1-2)
+
+**Do not write a single line of agent code yet.**
+
+Read `SKILL.md` → _Phase 0_. By the end you should be able to explain, in your
+own words and without looking:
+
+- Workflow vs Agent (and why building an agent when you need a workflow costs 10×)
+- The 5 workflow patterns (chaining, routing, parallelisation, orchestrator-worker, evaluator-optimiser)
+- The 4 context primitives (Write, Select, Compress, Isolate)
+- The harness stack (model = CPU, harness = OS)
+
+**Deliverable:** the 2-page concept doc. Use [`project/TEMPLATE.md`](project/TEMPLATE.md).
+If you can't write it without looking, you haven't understood it yet.

package/skills/craft/become-an-ai-engineer-26/phases/phase-0-mental-models/project/TEMPLATE.md

@@ -1,33 +1,33 @@
-# [Your Name] — AI Engineering Mental Models
-
-> 2-page concept doc. In your own words. No copy-paste.
-
-## 1. Workflow vs Agent
-
-- Workflow is...
-- Agent is...
-- I would choose an agent over a workflow when...
-
-## 2. The 5 Workflow Patterns
-
-1. Prompt chaining —
-2. Routing —
-3. Parallelisation —
-4. Orchestrator-worker —
-5. Evaluator-optimiser —
-
-## 3. The 4 Context Primitives
-
-- Write —
-- Select —
-- Compress —
-- Isolate —
-
-## 4. The Harness Stack
-
-- Why "the harness is the job":
-- One thing a harness gives the model that a raw API call does not:
-
-## 5. The one idea I understood only after writing this doc
-
--
+# [Your Name] — AI Engineering Mental Models
+
+> 2-page concept doc. In your own words. No copy-paste.
+
+## 1. Workflow vs Agent
+
+- Workflow is...
+- Agent is...
+- I would choose an agent over a workflow when...
+
+## 2. The 5 Workflow Patterns
+
+1. Prompt chaining —
+2. Routing —
+3. Parallelisation —
+4. Orchestrator-worker —
+5. Evaluator-optimiser —
+
+## 3. The 4 Context Primitives
+
+- Write —
+- Select —
+- Compress —
+- Isolate —
+
+## 4. The Harness Stack
+
+- Why "the harness is the job":
+- One thing a harness gives the model that a raw API call does not:
+
+## 5. The one idea I understood only after writing this doc
+
+-

package/skills/craft/become-an-ai-engineer-26/phases/phase-1-first-agent/README.md

@@ -1,25 +1,25 @@
-# Phase 1 — Build Your First Agent From Scratch (Weeks 3-5)
-
-Write the agent twice. Raw API first, Agent SDK second. Feel the difference.
-
-## Build #1 — The raw loop
-
-[`code/raw_loop.py`](code/raw_loop.py) — ~100 lines, no framework. Give it three
-tools (`web_search`, `read_file`, `write_file`), run it on a research task, and
-read every step of the trace.
-
-```bash
-cd code
-pip install anthropic
-export ANTHROPIC_API_KEY=...      # PowerShell: $env:ANTHROPIC_API_KEY="..."
-python raw_loop.py "Research the top 3 AI agent frameworks in 2026"
-```
-
-## Build #2 — The same agent on Claude Agent SDK
-
-Add `CLAUDE.md`, a `skills/research/SKILL.md`, a `hooks/post_tool.py` PostToolUse
-hook, and spawn one sub-agent via the Task tool. See `SKILL.md` → _Phase 1_.
-
-**Deliverable:** a daily-briefing agent. Reads your notes + RSS, writes a
-summarised briefing to disk every morning. Run it for a week. Watch it fail.
-Fix it.
+# Phase 1 — Build Your First Agent From Scratch (Weeks 3-5)
+
+Write the agent twice. Raw API first, Agent SDK second. Feel the difference.
+
+## Build #1 — The raw loop
+
+[`code/raw_loop.py`](code/raw_loop.py) — ~100 lines, no framework. Give it three
+tools (`web_search`, `read_file`, `write_file`), run it on a research task, and
+read every step of the trace.
+
+```bash
+cd code
+pip install anthropic
+export ANTHROPIC_API_KEY=...      # PowerShell: $env:ANTHROPIC_API_KEY="..."
+python raw_loop.py "Research the top 3 AI agent frameworks in 2026"
+```
+
+## Build #2 — The same agent on Claude Agent SDK
+
+Add `CLAUDE.md`, a `skills/research/SKILL.md`, a `hooks/post_tool.py` PostToolUse
+hook, and spawn one sub-agent via the Task tool. See `SKILL.md` → _Phase 1_.
+
+**Deliverable:** a daily-briefing agent. Reads your notes + RSS, writes a
+summarised briefing to disk every morning. Run it for a week. Watch it fail.
+Fix it.

package/skills/craft/become-an-ai-engineer-26/phases/phase-1-first-agent/code/raw_loop.py

@@ -1,126 +1,126 @@
-"""
-Phase 1 — The Raw Agent Loop (~100 lines)
-
-The entire job of a harness, stripped to its skeleton. No framework.
-Just the loop: the model thinks, picks a tool, you run it, you feed the
-result back, repeat until the model says it's done.
-
-Build this first. Read every step of the trace. Once you can see what the
-loop does for you, you'll understand what every framework is hiding.
-
-Requires:  pip install anthropic
-Env:       ANTHROPIC_API_KEY
-Run:       python raw_loop.py "Research the top 3 AI agent frameworks in 2026"
-"""
-
-import json
-import sys
-from anthropic import Anthropic
-
-client = Anthropic()  # reads ANTHROPIC_API_KEY from the environment
-
-# --- 1. Define the tools the model is allowed to call -----------------------
-TOOLS = [
-    {
-        "name": "web_search",
-        "description": "Search the web and return short text snippets.",
-        "input_schema": {
-            "type": "object",
-            "properties": {"query": {"type": "string"}},
-            "required": ["query"],
-        },
-    },
-    {
-        "name": "read_file",
-        "description": "Read a UTF-8 text file from disk and return its contents.",
-        "input_schema": {
-            "type": "object",
-            "properties": {"path": {"type": "string"}},
-            "required": ["path"],
-        },
-    },
-    {
-        "name": "write_file",
-        "description": "Write text to a file on disk. Overwrites if it exists.",
-        "input_schema": {
-            "type": "object",
-            "properties": {
-                "path": {"type": "string"},
-                "content": {"type": "string"},
-            },
-            "required": ["path", "content"],
-        },
-    },
-]
-
-
-# --- 2. Implement the tools -------------------------------------------------
-def web_search(query: str) -> str:
-    # Stub. Wire to a real search API (Tavily, Brave, SerpAPI) in practice.
-    return f"[search results for: {query!r}] — replace this stub with a real API."
-
-
-def read_file(path: str) -> str:
-    try:
-        with open(path, "r", encoding="utf-8") as f:
-            return f.read()
-    except Exception as e:  # surface the error to the model, don't crash the loop
-        return f"ERROR reading {path}: {e}"
-
-
-def write_file(path: str, content: str) -> str:
-    try:
-        with open(path, "w", encoding="utf-8") as f:
-            f.write(content)
-        return f"Wrote {len(content)} chars to {path}."
-    except Exception as e:
-        return f"ERROR writing {path}: {e}"
-
-
-def execute_tool(name: str, args: dict) -> str:
-    return {
-        "web_search": web_search,
-        "read_file": read_file,
-        "write_file": write_file,
-    }[name](**args)
-
-
-# --- 3. The loop ------------------------------------------------------------
-def run(user_input: str) -> None:
-    messages = [{"role": "user", "content": user_input}]
-    stop_reason = None
-
-    while stop_reason != "end_turn":
-        response = client.messages.create(
-            model="claude-sonnet-4-5",
-            max_tokens=2048,
-            messages=messages,
-            tools=TOOLS,
-        )
-        stop_reason = response.stop_reason
-
-        # Echo any text the model produced this turn
-        for block in response.content:
-            if block.type == "text":
-                print(block.text)
-
-        if stop_reason == "tool_use":
-            tool_results = []
-            for block in response.content:
-                if block.type == "tool_use":
-                    print(f"  -> {block.name}({json.dumps(block.input)})")
-                    result = execute_tool(block.name, block.input)
-                    tool_results.append(
-                        {
-                            "type": "tool_result",
-                            "tool_use_id": block.id,
-                            "content": result,
-                        }
-                    )
-            messages.append({"role": "assistant", "content": response.content})
-            messages.append({"role": "user", "content": tool_results})
-
-
-if __name__ == "__main__":
-    task = sys.argv[1] if len(sys.argv) > 1 else "Say hello and stop."
-    run(task)
+"""
+Phase 1 — The Raw Agent Loop (~100 lines)
+
+The entire job of a harness, stripped to its skeleton. No framework.
+Just the loop: the model thinks, picks a tool, you run it, you feed the
+result back, repeat until the model says it's done.
+
+Build this first. Read every step of the trace. Once you can see what the
+loop does for you, you'll understand what every framework is hiding.
+
+Requires:  pip install anthropic
+Env:       ANTHROPIC_API_KEY
+Run:       python raw_loop.py "Research the top 3 AI agent frameworks in 2026"
+"""
+
+import json
+import sys
+from anthropic import Anthropic
+
+client = Anthropic()  # reads ANTHROPIC_API_KEY from the environment
+
+# --- 1. Define the tools the model is allowed to call -----------------------
+TOOLS = [
+    {
+        "name": "web_search",
+        "description": "Search the web and return short text snippets.",
+        "input_schema": {
+            "type": "object",
+            "properties": {"query": {"type": "string"}},
+            "required": ["query"],
+        },
+    },
+    {
+        "name": "read_file",
+        "description": "Read a UTF-8 text file from disk and return its contents.",
+        "input_schema": {
+            "type": "object",
+            "properties": {"path": {"type": "string"}},
+            "required": ["path"],
+        },
+    },
+    {
+        "name": "write_file",
+        "description": "Write text to a file on disk. Overwrites if it exists.",
+        "input_schema": {
+            "type": "object",
+            "properties": {
+                "path": {"type": "string"},
+                "content": {"type": "string"},
+            },
+            "required": ["path", "content"],
+        },
+    },
+]
+
+
+# --- 2. Implement the tools -------------------------------------------------
+def web_search(query: str) -> str:
+    # Stub. Wire to a real search API (Tavily, Brave, SerpAPI) in practice.
+    return f"[search results for: {query!r}] — replace this stub with a real API."
+
+
+def read_file(path: str) -> str:
+    try:
+        with open(path, "r", encoding="utf-8") as f:
+            return f.read()
+    except Exception as e:  # surface the error to the model, don't crash the loop
+        return f"ERROR reading {path}: {e}"
+
+
+def write_file(path: str, content: str) -> str:
+    try:
+        with open(path, "w", encoding="utf-8") as f:
+            f.write(content)
+        return f"Wrote {len(content)} chars to {path}."
+    except Exception as e:
+        return f"ERROR writing {path}: {e}"
+
+
+def execute_tool(name: str, args: dict) -> str:
+    return {
+        "web_search": web_search,
+        "read_file": read_file,
+        "write_file": write_file,
+    }[name](**args)
+
+
+# --- 3. The loop ------------------------------------------------------------
+def run(user_input: str) -> None:
+    messages = [{"role": "user", "content": user_input}]
+    stop_reason = None
+
+    while stop_reason != "end_turn":
+        response = client.messages.create(
+            model="claude-sonnet-4-5",
+            max_tokens=2048,
+            messages=messages,
+            tools=TOOLS,
+        )
+        stop_reason = response.stop_reason
+
+        # Echo any text the model produced this turn
+        for block in response.content:
+            if block.type == "text":
+                print(block.text)
+
+        if stop_reason == "tool_use":
+            tool_results = []
+            for block in response.content:
+                if block.type == "tool_use":
+                    print(f"  -> {block.name}({json.dumps(block.input)})")
+                    result = execute_tool(block.name, block.input)
+                    tool_results.append(
+                        {
+                            "type": "tool_result",
+                            "tool_use_id": block.id,
+                            "content": result,
+                        }
+                    )
+            messages.append({"role": "assistant", "content": response.content})
+            messages.append({"role": "user", "content": tool_results})
+
+
+if __name__ == "__main__":
+    task = sys.argv[1] if len(sys.argv) > 1 else "Say hello and stop."
+    run(task)

package/skills/craft/become-an-ai-engineer-26/phases/phase-2-architecture/README.md

@@ -1,17 +1,17 @@
-# Phase 2 — Real Architecture (Weeks 6-9)
-
-Build on **LangGraph + Deep Agents** — the production stack. See `SKILL.md` →
-_Phase 2_ for the full walkthrough (StateGraph, PostgresSaver checkpointing,
-middleware hooks, isolated-context sub-agents).
-
-**Deliverable:** a Research Analyst Agent.
-
-- Lead agent plans → spawns 3 search sub-agents in parallel (isolated context)
-- Sub-agents write results to files → a citation sub-agent verifies
-- Writer agent produces Markdown with inline citations
-- State persists via `PostgresSaver` (kill → resume)
-- Human-in-the-loop interrupt before exceeding $1 in tokens
-- Ship a LangSmith trace URL in your README
-
-> Reference implementation (`research_agent/graph.py`, `nodes.py`, `tools.py`)
-> is left for you to build — that's the point of the phase.
+# Phase 2 — Real Architecture (Weeks 6-9)
+
+Build on **LangGraph + Deep Agents** — the production stack. See `SKILL.md` →
+_Phase 2_ for the full walkthrough (StateGraph, PostgresSaver checkpointing,
+middleware hooks, isolated-context sub-agents).
+
+**Deliverable:** a Research Analyst Agent.
+
+- Lead agent plans → spawns 3 search sub-agents in parallel (isolated context)
+- Sub-agents write results to files → a citation sub-agent verifies
+- Writer agent produces Markdown with inline citations
+- State persists via `PostgresSaver` (kill → resume)
+- Human-in-the-loop interrupt before exceeding $1 in tokens
+- Ship a LangSmith trace URL in your README
+
+> Reference implementation (`research_agent/graph.py`, `nodes.py`, `tools.py`)
+> is left for you to build — that's the point of the phase.

package/skills/craft/become-an-ai-engineer-26/phases/phase-3-harness/README.md

@@ -1,17 +1,17 @@
-# Phase 3 — Build the Harness Yourself (Weeks 10-13)
-
-**The highest-leverage phase in the roadmap.** You will not make the right
-production trade-offs until you've built a harness once. See `SKILL.md` →
-_Phase 3_ for the 10 components.
-
-**The 10 components:** loop control · tool dispatch · context management
-(auto-compaction at 85%) · persistence · sub-agent orchestration · skills /
-progressive disclosure · hooks · observability (OpenTelemetry) · sandboxing ·
-auth brokering.
-
-**Deliverable:** a ~1,500-line mini-harness in Python, plus a 1,000-word
-post-mortem comparing it to the Claude Agent SDK and Deep Agents — what you got
-right, what you cut, what you'd do differently.
-
-> `code/mini_harness/` is yours to write. Don't copy a framework — build it,
-> then read the frameworks and see what they hid from you.
+# Phase 3 — Build the Harness Yourself (Weeks 10-13)
+
+**The highest-leverage phase in the roadmap.** You will not make the right
+production trade-offs until you've built a harness once. See `SKILL.md` →
+_Phase 3_ for the 10 components.
+
+**The 10 components:** loop control · tool dispatch · context management
+(auto-compaction at 85%) · persistence · sub-agent orchestration · skills /
+progressive disclosure · hooks · observability (OpenTelemetry) · sandboxing ·
+auth brokering.
+
+**Deliverable:** a ~1,500-line mini-harness in Python, plus a 1,000-word
+post-mortem comparing it to the Claude Agent SDK and Deep Agents — what you got
+right, what you cut, what you'd do differently.
+
+> `code/mini_harness/` is yours to write. Don't copy a framework — build it,
+> then read the frameworks and see what they hid from you.

package/skills/craft/become-an-ai-engineer-26/phases/phase-4-evals/README.md

@@ -1,21 +1,21 @@
-# Phase 4 — Evals & Regression Harness (Weeks 14-17)
-
-> Without this, every "improvement" is vibes.
-
-See `SKILL.md` → _Phase 4_. The four eval types: single-turn, trajectory,
-LLM-as-judge, end-state.
-
-The CI gate lives at
-[`code/.github/workflows/eval.yml`](code/.github/workflows/eval.yml) — a
-reference GitHub Actions workflow that blocks a merge if the golden-dataset
-pass rate drops 3+ points below baseline. Copy it into your own repo and wire
-`ANTHROPIC_API_KEY` as a repo secret.
-
-**Deliverable:**
-
-- Golden dataset: 30–50 hand-graded questions across 3 difficulty levels
-- Deterministic graders for factual queries
-- LLM-as-judge with a 5-criterion rubric
-- Trajectory eval: did it plan, spawn sub-agents, cite, stay under budget?
-- Wired into GitHub Actions, blocking merges on a 3-point regression
-- Production sampling: 1% of live traces auto-graded nightly
+# Phase 4 — Evals & Regression Harness (Weeks 14-17)
+
+> Without this, every "improvement" is vibes.
+
+See `SKILL.md` → _Phase 4_. The four eval types: single-turn, trajectory,
+LLM-as-judge, end-state.
+
+The CI gate lives at
+[`code/.github/workflows/eval.yml`](code/.github/workflows/eval.yml) — a
+reference GitHub Actions workflow that blocks a merge if the golden-dataset
+pass rate drops 3+ points below baseline. Copy it into your own repo and wire
+`ANTHROPIC_API_KEY` as a repo secret.
+
+**Deliverable:**
+
+- Golden dataset: 30–50 hand-graded questions across 3 difficulty levels
+- Deterministic graders for factual queries
+- LLM-as-judge with a 5-criterion rubric
+- Trajectory eval: did it plan, spawn sub-agents, cite, stay under budget?
+- Wired into GitHub Actions, blocking merges on a 3-point regression
+- Production sampling: 1% of live traces auto-graded nightly

package/skills/craft/become-an-ai-engineer-26/phases/phase-4-evals/code/.github/workflows/eval.yml

@@ -1,40 +1,40 @@
-# Agent Regression Harness — GitHub Actions
-#
-# Reference workflow from Phase 4. Copy into your own repo's
-# .github/workflows/ and wire ANTHROPIC_API_KEY as a repo secret.
-# Blocks the merge if the golden-dataset pass rate drops 3+ points
-# below the committed baseline.
-name: Agent Regression
-
-on: [pull_request]
-
-jobs:
-  eval:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-
-      - uses: actions/setup-python@v5
-        with:
-          python-version: "3.12"
-
-      - name: Install dependencies
-        run: pip install -r requirements.txt
-
-      - name: Run golden dataset
-        env:
-          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
-        run: python evals/run_golden.py
-
-      - name: Check pass rate against baseline
-        run: |
-          PASS_RATE=$(cat results/pass_rate.json | jq .rate)
-          BASELINE=$(cat evals/baseline.json | jq .rate)
-          python -c "
-          pass_rate = $PASS_RATE
-          baseline = $BASELINE
-          if pass_rate < baseline - 0.03:
-              print(f'FAIL: {pass_rate:.1%} < baseline {baseline:.1%} - 3pp')
-              exit(1)
-          print(f'PASS: {pass_rate:.1%} (baseline {baseline:.1%})')
-          "
+# Agent Regression Harness — GitHub Actions
+#
+# Reference workflow from Phase 4. Copy into your own repo's
+# .github/workflows/ and wire ANTHROPIC_API_KEY as a repo secret.
+# Blocks the merge if the golden-dataset pass rate drops 3+ points
+# below the committed baseline.
+name: Agent Regression
+
+on: [pull_request]
+
+jobs:
+  eval:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install dependencies
+        run: pip install -r requirements.txt
+
+      - name: Run golden dataset
+        env:
+          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
+        run: python evals/run_golden.py
+
+      - name: Check pass rate against baseline
+        run: |
+          PASS_RATE=$(cat results/pass_rate.json | jq .rate)
+          BASELINE=$(cat evals/baseline.json | jq .rate)
+          python -c "
+          pass_rate = $PASS_RATE
+          baseline = $BASELINE
+          if pass_rate < baseline - 0.03:
+              print(f'FAIL: {pass_rate:.1%} < baseline {baseline:.1%} - 3pp')
+              exit(1)
+          print(f'PASS: {pass_rate:.1%} (baseline {baseline:.1%})')
+          "