quick-agent 0.1.1__py3-none-any.whl → 0.1.3__py3-none-any.whl

{quick_agent-0.1.1.data → quick_agent-0.1.3.data}/data/quick_agent/agents/function-spec-validator.md

@@ -43,7 +43,7 @@ handoff:
   input_mode: "final_output_json"
 ---
 
-# function_spec_validator
+## Instructions
 
 You are a validation agent that checks a provided function specification.
 Return a JSON valid result that includes a boolean `valid` field.

{quick_agent-0.1.1.data → quick_agent-0.1.3.data}/data/quick_agent/agents/subagent-validate-eval-list.md

@@ -49,7 +49,7 @@ handoff:
   input_mode: "final_output_json"
 ---
 
-# subagent-validate-eval-list
+## Instructions
 
 You are an EVAL LIST EXECUTOR. Your sole responsibility is reading an eval list file, executing each test agent, validating responses, and producing a results summary. You do NOT perform the tasks yourself—you delegate them entirely.
 

{quick_agent-0.1.1.data → quick_agent-0.1.3.data}/data/quick_agent/agents/subagent-validator-contains.md

@@ -47,7 +47,7 @@ handoff:
   input_mode: "final_output_json"
 ---
 
-# subagent-validator-contains
+# Instructions
 
 You are a RESPONSE VALIDATOR. Your sole responsibility is comparing a response against expected text patterns defined in an eval.md file. You report PASS or FAIL with detailed results.
 
@@ -56,10 +56,12 @@ You are a RESPONSE VALIDATOR. Your sole responsibility is comparing a response a
 ### Step 1: Parse Input
 
 Extract from the user's request:
+
 - Response Text: the text to validate from a provided file path
 - Eval File Path: path to the markdown file containing expected text patterns
 
 Input formats accepted:
+
 - `validate "{response-file.md}" against {path/to/eval.md}`
 - `check response in {response-file.md} contains {eval.md}`
 - Direct response text followed by eval file path
@@ -86,26 +88,31 @@ Output a structured validation report and a PASS/FAIL status.
 ## step:plan
 
 Goal:
+
 - Identify response path/text and eval file path from the input.
 - Outline the minimal steps.
 
 Constraints:
+
 - Keep it short.
 
 ## step:execute
 
 Goal:
+
 - Read the response file (if a path is provided).
 - Read the eval file.
 - Check for each expected text line in the response.
 - Note missing lines.
 
 Constraints:
+
 - Do not output JSON in this step.
 
 ## step:finalize
 
 Goal:
+
 - Return a `ContainsValidationResult` JSON object with:
   - `status`: PASS if all expected lines are present, otherwise FAIL
   - `checks`: list of per-line results

{quick_agent-0.1.1.data → quick_agent-0.1.3.data}/data/quick_agent/agents/template.md

@@ -52,7 +52,13 @@ handoff:
   input_mode: "final_output_json"   # or "final_output_markdown"
 ---
 
-# doc_pipeline_agent
+# System Prompt
+
+This is a system prompt to included in every run
+
+## Instructions
+
+Instructions are only included in first run.
 
 You are a reliable pipeline agent.
 You must follow the chain steps in order.
@@ -61,26 +67,31 @@ You may call tools as needed. If you call `agent.call`, wait for the response an
 ## step:plan
 
 Goal:
+
 - Read the provided input (a JSON or Markdown/text file) embedded by the orchestrator.
 - Produce a structured **Plan** that lists concrete actions and any tool calls required.
 
 Constraints:
+
 - Keep steps explicit.
 - If you need another agent, call `agent.call` with a clear request.
 
 ## step:execute
 
 Goal:
+
 - Execute the plan.
 - Use the declared tools. You may call tools multiple times.
 
 Constraints:
+
 - Write intermediate artifacts only if asked.
 - Summarize what you did in plain text.
 
 ## step:finalize
 
 Goal:
+
 - Produce a final **FinalResult** object that is valid JSON for the schema.
 - Include references to tools invoked and any sub-agent calls.
 - If anything failed, reflect it in the structured fields rather than “hiding” it in prose.

{quick_agent-0.1.1.dist-info → quick_agent-0.1.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: quick-agent
-Version: 0.1.1
+Version: 0.1.3
 Summary: Minimal, local-first agent runner using Markdown front matter.
 Author-email: Charles Verge <1906614+charlesverge@users.noreply.github.com>
 License:                     GNU GENERAL PUBLIC LICENSE
@@ -700,6 +700,8 @@ Requires-Dist: anyio
 Requires-Dist: pydantic
 Requires-Dist: pydantic-ai
 Requires-Dist: python-frontmatter
+Requires-Dist: types-PyYAML
+Requires-Dist: PyYAML
 Provides-Extra: dev
 Requires-Dist: mypy; extra == "dev"
 Requires-Dist: ruff; extra == "dev"
@@ -707,13 +709,14 @@ Provides-Extra: test
 Requires-Dist: pytest; extra == "test"
 Dynamic: license-file
 
-# Simple Agent
+# Quick Agent
 
-Simple Agent is a minimal, local-first agent runner that loads agent definitions from Markdown front matter and executes a small chain of steps with limited context handling. It is intentionally small and explicit: you define the model, tools, and steps in a single Markdown file, and the orchestrator runs those steps in order with a bounded prompt preamble.
+Quick Agent is a minimal, local-first agent runner that loads agent definitions from Markdown front matter and executes a small chain of steps with limited context handling. It is intentionally small and explicit: you define the model, tools, and steps in a single Markdown file, and the orchestrator runs those steps in order with a bounded prompt preamble.
 
 ## Project Goal
 
 Provide a simple, maintainable agent framework that:
+
 - Uses Markdown front matter for agent configuration.
 - Runs a deterministic chain of steps (text or structured output).
 - Keeps context handling deliberately limited and predictable.
@@ -887,7 +890,7 @@ def main() -> None:
         tools=tools,
         directory_permissions=permissions,
         agent_id="hello",
-        input_path=Path("safe/path/to/input.txt"),
+        input_data=Path("safe/path/to/input.txt"),
         extra_tools=None,
     )
 
@@ -907,6 +910,18 @@ Agents are stored as Markdown files with YAML front matter and step sections:
 - Body contains `## step:<id>` sections referenced by the chain.
 
 The orchestrator loads the agent, builds the tools, and executes each step in order, writing the final output to disk.
+If the agent front matter omits `output.file`, the orchestrator returns the final output without writing a file.
+
+## Nested Output
+
+When an agent invokes another agent via `agent_call` or `handoff`, the nested agent can either write its
+own `output.file` or return output inline only. Configure this in the parent agent front matter:
+
+```yaml
+nested_output: inline  # default, no output file for nested calls
+```
+
+Use `nested_output: file` to allow nested agents to write their configured output files.
 
 ## Documentation
 
@@ -914,5 +929,7 @@ See the docs in `docs/`:
 
 - [docs/cli.md](docs/cli.md): Command line usage and options.
 - [docs/templates.md](docs/templates.md): Agent template format and examples.
+- [docs/outputs.md](docs/outputs.md): Output configuration and behavior.
 - [docs/python.md](docs/python.md): Embedding the orchestrator in scripts.
 - [docs/python.md#inter-agent-calls](docs/python.md#inter-agent-calls): Example of one agent calling another.
+- [src/quick_agent/llms.txt](src/quick_agent/llms.txt): LLM-oriented project summary and examples.

quick_agent-0.1.3.dist-info/RECORD

@@ -0,0 +1,52 @@
+quick_agent/__init__.py,sha256=lgTMjPhbumIZ1Pna33khCaAs4DSjErnYoCkvK_YxBCc,362
+quick_agent/agent_call_tool.py,sha256=yyXpWWpQlhSV8qxkeA1tLdrfcHhYg2cmhf4sxx8FfvA,2195
+quick_agent/agent_registry.py,sha256=VqDFYsH6-kwUwtuiQyHpvmDeWVcSbn62pCvqE8D7IZE,1818
+quick_agent/agent_tools.py,sha256=GwaPkoG_vy1f6hatvRQFoDj0IJ-6nIYV746qc8y-yMg,1475
+quick_agent/cli.py,sha256=nzVF4paXxugaPzmHTMwCBkbQkfkkv9hqGQH4Fh0jJ_U,1973
+quick_agent/directory_permissions.py,sha256=r6Lc56oIrGhVVUVrCoPInue_BCdGZsD5oDgoAHJWiyE,1955
+quick_agent/input_adaptors.py,sha256=ZaK0Mm27cGMGb5y6fsgV0gXRPMW2zmHn9l9dXKzOufA,865
+quick_agent/io_utils.py,sha256=BUmUoZepL4lSYe0JbKxLm4mFFQ6Zt6MZQzprzgoBweU,1200
+quick_agent/json_utils.py,sha256=G6uP9SrdEw5WtA9--dBqdJcTYD2JgdgotPJdleErB_c,1003
+quick_agent/llms.txt,sha256=_jJvgJDdGyuIc-sT3XoadCGxDhzi61Cnb3Fq6zpsE_M,6488
+quick_agent/orchestrator.py,sha256=SGa5oeRgbjkKXdFL5kszZT-R6-jIgyOkHEMmjLrDWjc,1321
+quick_agent/prompting.py,sha256=ZxKZ-qKfbStj7160xi84RYJWB5BxhYD01-ruiTN4iQA,1317
+quick_agent/py.typed,sha256=3Q9vdii9v-_pbj9sglgDzJAIelvARaYaDJlB2u-NDqg,38
+quick_agent/quick_agent.py,sha256=P5qlJrdTwwYAafkm_ML5DJT3w49QRtEvUpF3fiJjR6Q,12850
+quick_agent/tools_loader.py,sha256=oveR-EGAZo3Q7NPJ2aWKU3d8xd8kIpibOrjInpUZwQg,2642
+quick_agent/models/__init__.py,sha256=YseFAVWLarh0wZCVUL0fmkjVgA-FnYJLEiVinG1-6Tk,737
+quick_agent/models/agent_spec.py,sha256=YsCrs9DSKJJJBsqnMAxb_19zG6UuNHtXwiuTZjbFg5U,882
+quick_agent/models/chain_step_spec.py,sha256=NcdYFmgJhA5ODIwCTgTb2fmmJJesAZCVAIy3cGC5dhs,342
+quick_agent/models/handoff_spec.py,sha256=46Pt4JGU8-6f-cdbDR0oqWuJbg5ENJcXrFN1IK3chvs,280
+quick_agent/models/loaded_agent_file.py,sha256=kOOrzfoYqCgoifcnOtn2us_MNPSVg4kRWgmyeb21BEY,5159
+quick_agent/models/model_spec.py,sha256=spnIZ8BNtXT5t8YRL1nh7cXrgWwNSptt_BMnAB8w1Lg,425
+quick_agent/models/output_spec.py,sha256=ZGYmHXAh82SoD5g8VPJRu_2uTjL_W_NM0gCXrAfiYYI,223
+quick_agent/models/run_input.py,sha256=WpV-QPlOPXMmZR6eIvP7vJOuCeBt3xWe4j4ngEbmSxg,282
+quick_agent/models/tool_impl_spec.py,sha256=7B161m8nFZCNjslb4VZdSi-mdAZM0jhp4kV4uaq8X1s,216
+quick_agent/models/tool_json.py,sha256=eYFBPCqxmRfyo7GYHBXCfog7kUzpjginfjt9grQy4ic,274
+quick_agent/schemas/outputs.py,sha256=8FTIczUV079imbygbFU3h9tyLWSH8aQZCPBTxM8QVtk,1527
+quick_agent/tools/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+quick_agent/tools/filesystem/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+quick_agent/tools/filesystem/adapter.py,sha256=aYcX1qjI9wRSANSNSB3KWbKi7uDIYo3EpJotpUxnQUM,1027
+quick_agent/tools/filesystem/read_text.py,sha256=owpZxwCFoYbljABhwpNKU4g7-8xDPJDu0kZYF7bPLfs,478
+quick_agent/tools/filesystem/write_text.py,sha256=lwsMCk_p8xurVxychM2UBieXDp0SgN3C6rufLBUzDxo,584
+quick_agent/tools/filesystem.read_text/tool.json,sha256=UtkOxcpi6qYd611tJx-zFXVUjOwx2ZbmxEWsF4cAVpo,246
+quick_agent/tools/filesystem.write_text/tool.json,sha256=EcGmB_M_KDTqOrnxk_NVhKRoW_KADFFksgJ7fmw472s,275
+quick_agent-0.1.3.data/data/quick_agent/agents/business-extract-structured.md,sha256=yeNIZFw60-2u93ZdI6nnQnTOGrllCN8SQLJPApvqXdU,1168
+quick_agent-0.1.3.data/data/quick_agent/agents/business-extract.md,sha256=SZAgpRI8Z4HnkMBAwvMLrErkNQ5HkHqerVMSisn-gNU,993
+quick_agent-0.1.3.data/data/quick_agent/agents/function-spec-validator.md,sha256=ls3r0ejswgM_9SljMLxMjk75x9XJna7ceqHX3pWcgnk,3279
+quick_agent-0.1.3.data/data/quick_agent/agents/subagent-validate-eval-list.md,sha256=IxFwHGIsnGbnaCK-j7znrp4llShepRfRlhTJ5sjqe-g,4105
+quick_agent-0.1.3.data/data/quick_agent/agents/subagent-validator-contains.md,sha256=1yY3Z5copRoCIvq_taVSjzvp8prFab49s1-ybaDcdpk,3107
+quick_agent-0.1.3.data/data/quick_agent/agents/template.md,sha256=4_ldsHj46qmDcer4SCfwfxRJqoesZsp87XHSAimtaLo,2571
+quick_agent-0.1.3.dist-info/licenses/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
+tests/test_agent.py,sha256=1FQteGMCftv40bC-i_iNETaLR-7hmUz9kIlcJYvnazU,10999
+tests/test_directory_permissions.py,sha256=-9h7W1VO0LekPLCxEqIAGgebquNks2F8M02BdZDLlQY,2815
+tests/test_httpx_tools.py,sha256=xqIQ376STyNGwd5YA8VZPZS4nsr1EsSDYCrrcoOuXBE,10500
+tests/test_input_adaptors.py,sha256=sle9zrumq3CUVGtDPZi2pf0om-Z7v4SI6khnTf7Rz4c,918
+tests/test_integration.py,sha256=XbI6abGeoZm3r_BqpQyhxhF36N7VYtNGcJ9-xNxZysM,9044
+tests/test_orchestrator.py,sha256=Tc1iynzPMf_ghbHitlcKDS9dks91t-ojbG3l7qInsbA,41716
+tests/test_tools.py,sha256=Yc6AKCz79XJwHqiRV8dUBE6GVNdspUU0hIeGCXx-Rvc,936
+quick_agent-0.1.3.dist-info/METADATA,sha256=x_ButLyAVP8hACb2fwZsHzPAQbYQQ6Li-TgsUP55hpE,47339
+quick_agent-0.1.3.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+quick_agent-0.1.3.dist-info/entry_points.txt,sha256=bij-xFaQMSrMj7zearzi-bMfcyweum_GvURFMVZGde8,53
+quick_agent-0.1.3.dist-info/top_level.txt,sha256=KT1ID0FVC0OLzQnoBKRIHbXLRNugiU2gcgj_f4DtAsw,18
+quick_agent-0.1.3.dist-info/RECORD,,

tests/test_agent.py

@@ -1,4 +1,5 @@
 import json
+import logging
 import sys
 import types
 from pathlib import Path
@@ -79,16 +80,24 @@ def test_write_output_creates_parent(tmp_path: Path) -> None:
 
 def test_make_user_prompt_contains_sections() -> None:
     run_input = RunInput(source_path="file.txt", kind="text", text="hi", data=None)
-    prompt = prompting.make_user_prompt("do thing", run_input, {"x": 1})
+    prompt = prompting.make_user_prompt(run_input, {"steps": {"x": 1}})
 
     assert "# Task Input" in prompt
     assert "source_path: file.txt" in prompt
     assert "## Input Content" in prompt
     assert "hi" in prompt
-    assert "## Chain State (JSON)" in prompt
-    assert '"x": 1' in prompt
-    assert "## Step Instructions" in prompt
-    assert "do thing" in prompt
+    assert "## Chain State (YAML)" in prompt
+    assert "x: 1" in prompt
+
+
+def test_make_user_prompt_inline_without_state_hides_headers() -> None:
+    run_input = RunInput(source_path="inline_input.txt", kind="text", text="hello", data=None)
+    prompt = prompting.make_user_prompt(run_input, {})
+
+    assert "# Task Input" not in prompt
+    assert "## Input Content" not in prompt
+    assert "## Chain State" not in prompt
+    assert prompt.strip() == "hello"
 
 
 def test_resolve_schema_valid_missing_and_invalid() -> None:
@@ -110,7 +119,7 @@ def test_resolve_schema_valid_missing_and_invalid() -> None:
         chain=[ChainStepSpec(id="s1", kind="text", prompt_section="step:one")],
         schemas={"Good": "schemas.tmp:GoodSchema", "Bad": "schemas.tmp:NotSchema"},
     )
-    loaded = LoadedAgentFile(spec=spec, body="", step_prompts={})
+    loaded = LoadedAgentFile.from_parts(spec=spec, instructions="", system_prompt="", step_prompts={})
 
     try:
         assert resolve_schema(loaded, "Good") is GoodSchema
@@ -137,6 +146,10 @@ chain:
     prompt_section: step:one
 ---
 
+## Instructions
+
+system.
+
 ## step:one
 
 body.
@@ -144,8 +157,9 @@ body.
     md_path = tmp_path / "agent.md"
     md_path.write_text(md, encoding="utf-8")
 
-    loaded = agent_registry.load_agent_file(md_path)
+    loaded = LoadedAgentFile(md_path)
     assert loaded.spec.name == "Test"
+    assert loaded.instructions == "system."
     assert loaded.step_prompts["step:one"] == "body."
 
 
@@ -166,7 +180,7 @@ body.
     md_path = tmp_path / "agent.md"
     md_path.write_text(md, encoding="utf-8")
 
-    loaded = agent_registry.load_agent_file(md_path)
+    loaded = LoadedAgentFile(md_path)
     assert loaded.spec.model.base_url == "https://api.openai.com/v1"
     assert loaded.spec.model.api_key_env == "OPENAI_API_KEY"
     assert loaded.spec.model.model_name == "gpt-5.2"
@@ -189,8 +203,258 @@ body.
     md_path = tmp_path / "agent.md"
     md_path.write_text(md, encoding="utf-8")
 
-    loaded = agent_registry.load_agent_file(md_path)
+    loaded = LoadedAgentFile(md_path)
     assert loaded.spec.model.base_url == "https://api.openai.com/v1"
     assert loaded.spec.model.api_key_env == "OPENAI_API_KEY"
     assert loaded.spec.model.model_name == "gpt-5.2"
     assert loaded.spec.model.provider == "openai-compatible"
+
+
+def test_load_agent_file_parses_instructions_and_system_prompt(tmp_path: Path) -> None:
+    md = """---
+name: Sections
+model:
+  base_url: http://localhost
+  model_name: test
+chain: []
+---
+
+## instructions
+
+Use the tool.
+
+## Notes
+
+ignored.
+
+## System prompt
+
+You are concise.
+"""
+    md_path = tmp_path / "agent.md"
+    md_path.write_text(md, encoding="utf-8")
+
+    loaded = LoadedAgentFile(md_path)
+    assert "Use the tool." in loaded.instructions
+    assert "## Notes" in loaded.instructions
+    assert "ignored." in loaded.instructions
+    assert loaded.system_prompt == "You are concise."
+    assert loaded.step_prompts == {}
+
+
+@pytest.mark.parametrize(
+    ("header", "expected"),
+    [
+        ("## Instructions", "Do it."),
+        ("## instructions", "Do it."),
+        ("## INSTRUCTIONS", "Do it."),
+    ],
+)
+def test_load_agent_file_accepts_instruction_header_case(
+    tmp_path: Path, header: str, expected: str
+) -> None:
+    md = f"""---
+name: Case Instructions
+model:
+  base_url: http://localhost
+  model_name: test
+chain: []
+---
+
+{header}
+
+{expected}
+"""
+    md_path = tmp_path / "agent.md"
+    md_path.write_text(md, encoding="utf-8")
+
+    loaded = LoadedAgentFile(md_path)
+    assert loaded.instructions == expected
+
+
+@pytest.mark.parametrize(
+    ("header", "expected"),
+    [
+        ("# System prompt", "Be brief."),
+        ("## System prompt", "Be brief."),
+        ("### System prompt", "Be brief."),
+        ("## system prompt", "Be brief."),
+        ("## SYSTEM PROMPT", "Be brief."),
+        ("## system_prompt", "Be brief."),
+    ],
+)
+def test_load_agent_file_accepts_system_prompt_header_case(
+    tmp_path: Path, header: str, expected: str
+) -> None:
+    md = f"""---
+name: Case System
+model:
+  base_url: http://localhost
+  model_name: test
+chain: []
+---
+
+{header}
+
+{expected}
+"""
+    md_path = tmp_path / "agent.md"
+    md_path.write_text(md, encoding="utf-8")
+
+    loaded = LoadedAgentFile(md_path)
+    assert loaded.system_prompt == expected
+
+
+@pytest.mark.parametrize(
+    "header",
+    [
+        "## step:one",
+        "## STEP:one",
+        "## Step:one",
+    ],
+)
+def test_load_agent_file_accepts_step_header_case(tmp_path: Path, header: str) -> None:
+    md = f"""---
+name: Case Step
+model:
+  base_url: http://localhost
+  model_name: test
+chain:
+  - id: one
+    kind: text
+    prompt_section: step:one
+---
+
+{header}
+
+Say hi.
+"""
+    md_path = tmp_path / "agent.md"
+    md_path.write_text(md, encoding="utf-8")
+
+    loaded = LoadedAgentFile(md_path)
+    assert loaded.step_prompts["step:one"] == "Say hi."
+
+
+def test_load_agent_file_allows_instructions_only_no_steps(tmp_path: Path) -> None:
+    md = """---
+name: No Steps
+model:
+  base_url: http://localhost
+  model_name: test
+chain: []
+---
+
+## Instructions
+
+Just answer.
+"""
+    md_path = tmp_path / "agent.md"
+    md_path.write_text(md, encoding="utf-8")
+
+    loaded = LoadedAgentFile(md_path)
+    assert loaded.instructions == "Just answer."
+    assert loaded.step_prompts == {}
+
+
+def test_load_agent_file_raises_without_sections(tmp_path: Path) -> None:
+    md = """---
+name: Empty
+model:
+  base_url: http://localhost
+  model_name: test
+chain: []
+---
+
+just text.
+"""
+    md_path = tmp_path / "agent.md"
+    md_path.write_text(md, encoding="utf-8")
+
+    with pytest.raises(ValueError, match="instructions, system prompt, or step sections"):
+        LoadedAgentFile(md_path)
+
+
+def test_load_agent_file_warns_on_preamble_before_instructions(tmp_path: Path, caplog: pytest.LogCaptureFixture) -> None:
+    md = """---
+name: Preamble
+model:
+  base_url: http://localhost
+  model_name: test
+chain: []
+---
+
+Preamble text.
+
+## Instructions
+
+Do the thing.
+"""
+    md_path = tmp_path / "agent.md"
+    md_path.write_text(md, encoding="utf-8")
+
+    with caplog.at_level(logging.WARNING):
+        LoadedAgentFile(md_path)
+
+    assert "Ignored text before instructions or system prompt" in caplog.text
+
+
+def test_load_agent_file_preserves_subsections_in_all_sections(tmp_path: Path) -> None:
+    md = """---
+name: Subsections
+model:
+  base_url: http://localhost
+  model_name: test
+chain:
+  - id: one
+    kind: text
+    prompt_section: step:one
+---
+
+## Instructions
+
+Intro line.
+
+## Constraints
+
+- Be concise.
+- Keep scope tight.
+
+## System prompt
+
+System intro.
+
+## Rules
+
+Always follow the rules.
+
+## step:one
+
+Step intro.
+
+## Details
+
+Explain details here.
+
+## step:two
+
+Step intro two.
+
+## Details
+
+Explain details here.
+"""
+    md_path = tmp_path / "agent.md"
+    md_path.write_text(md, encoding="utf-8")
+
+    loaded = LoadedAgentFile(md_path)
+    assert "Intro line." in loaded.instructions
+    assert "## Constraints" in loaded.instructions
+    assert "Be concise." in loaded.instructions
+    assert "System intro." in loaded.system_prompt
+    assert "## Rules" in loaded.system_prompt
+    assert "Always follow the rules." in loaded.system_prompt
+    assert "Step intro." in loaded.step_prompts["step:one"]
+    assert "## Details" in loaded.step_prompts["step:one"]
+    assert "Explain details here." in loaded.step_prompts["step:one"]
+    assert "## Details" in loaded.step_prompts["step:two"]

tests/test_directory_permissions.py

@@ -45,6 +45,16 @@ def test_directory_permissions_can_read_write(tmp_path: Path) -> None:
     assert perms.can_write(Path("../nope.txt")) is False
 
 
+def test_directory_permissions_without_root_denies_all() -> None:
+    perms = DirectoryPermissions(None)
+
+    with pytest.raises(PermissionError):
+        perms.resolve(Path("anything.txt"), for_write=False)
+
+    assert perms.can_read(Path("anything.txt")) is False
+    assert perms.can_write(Path("anything.txt")) is False
+
+
 def test_agent_cannot_write_outside_scoped_directory(
     tmp_path: Path,
     monkeypatch: pytest.MonkeyPatch,