agentic-dev-loop 1.0.0a37__tar.gz → 1.0.0a38__tar.gz

{agentic_dev_loop-1.0.0a37/src/agentic_dev_loop.egg-info → agentic_dev_loop-1.0.0a38}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agentic-dev-loop
-Version: 1.0.0a37
+Version: 1.0.0a38
 Summary: Turn an LLM into a gated, sandboxed, resumable construction loop over your codebase.
 Author: Luke Wilson
 License-Expression: Apache-2.0

{agentic_dev_loop-1.0.0a37 → agentic_dev_loop-1.0.0a38}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "agentic-dev-loop"
-version = "1.0.0a37"
+version = "1.0.0a38"
 description = "Turn an LLM into a gated, sandboxed, resumable construction loop over your codebase."
 readme = "README.md"
 requires-python = ">=3.12"

{agentic_dev_loop-1.0.0a37 → agentic_dev_loop-1.0.0a38}/src/agentic_dev_loop/core/config.py

@@ -389,6 +389,7 @@ class StepConfig(BaseModel):
     stop_when: StopWhenSpec | None = None  # agent steps only (032)
     finalize_nudge_turns: int | None = Field(default=None, gt=0)  # agent steps only (035)
     brief_inputs: BriefInputsSpec | None = None  # agent steps only (033)
+    custom_prompt: str | None = None  # agent steps only (040); file path resolved at load
     env: dict[str, str | None] | None = None  # NFR-1: null = inherit host value
     setup: str | None = None  # BR-13: command steps only
     forbidden_commands: list[str] = []  # banned-command regex deny-list (036 D1)
@@ -441,6 +442,13 @@ class StepConfig(BaseModel):
                     f"regex: {e}") from e
         return v
 
+    @field_validator("custom_prompt")
+    @classmethod
+    def _nonempty_custom_prompt(cls, v: str | None) -> str | None:
+        if v is not None and not v.strip():
+            raise ValueError("custom_prompt must be a non-empty string")
+        return v
+
     @model_validator(mode="after")
     def _check_shape(self) -> "StepConfig":
         for key in ("agent", "action", "command"):
@@ -501,6 +509,10 @@ class StepConfig(BaseModel):
             raise ValueError(
                 f"step {self.name!r}: 'brief_inputs' requires 'agent'"
             )
+        if self.custom_prompt is not None and self.agent is None:
+            raise ValueError(
+                f"step {self.name!r}: 'custom_prompt' requires 'agent'"
+            )
         if (self.brief_inputs is not None
                 and (self.artifact is not None or self.verdicts is not None)):
             raise ValueError(
@@ -797,6 +809,28 @@ class PipelineConfig(BaseModel):
         return self
 
 
+def _custom_prompt_looks_like_path(value: str) -> bool:
+    """040/D2: a value counts as path-shaped when it carries a path
+    separator or a file extension (a dot). A path-shaped value that names no
+    existing file is a hard error, never silent text."""
+    return "/" in value or "\\" in value or Path(value).suffix != ""
+
+
+def resolve_custom_prompt(value: str, pipeline_dir: Path, step_name: str) -> str:
+    """Resolve a ``custom_prompt`` value (040/D2): an existing file
+    (relative to the pipeline's directory, or absolute) is read as the
+    prompt; any other value is literal text. A path-shaped value that names
+    no file fails loud, naming the path."""
+    candidate = Path(value)
+    path = candidate if candidate.is_absolute() else pipeline_dir / candidate
+    if path.is_file():
+        return path.read_text(encoding="utf-8")
+    if _custom_prompt_looks_like_path(value):
+        raise ValueError(
+            f"step {step_name!r}: 'custom_prompt' file not found: {value}")
+    return value
+
+
 def default_state_dir(pipeline_path: Path) -> str:
     """The new run-state default (Q1): checkpoint DB + last-error.md live OUTSIDE
     the workspace mount, under ~/.agentic-dev-loop/runs/<pipeline-hash>/ where
@@ -816,4 +850,10 @@ def load_config(path: Path) -> PipelineConfig:
         )
     if "state_dir" not in data:
         data["state_dir"] = default_state_dir(path)
-    return PipelineConfig.model_validate(data)
+    cfg = PipelineConfig.model_validate(data)
+    pipeline_dir = path.resolve().parent
+    for step in cfg.steps:
+        if step.custom_prompt is not None:
+            step.custom_prompt = resolve_custom_prompt(
+                step.custom_prompt, pipeline_dir, step.name)
+    return cfg

{agentic_dev_loop-1.0.0a37 → agentic_dev_loop-1.0.0a38}/src/agentic_dev_loop/engine/stages.py

@@ -58,6 +58,13 @@ DEFAULT_MAX_ITERATIONS = 500
 DEFAULT_COMMAND_TIMEOUT = 600.0  # seconds - a pytest-class suite budget
 LAST_ERROR = "last-error.md"
 ARTIFACT_CONTEXT_CAP = 24_000  # chars per artifact injected into a step's context
+CUSTOM_PROMPT_HEADING = "Run-specific instructions"
+
+
+def custom_prompt_section(text: str) -> str:
+    """The per-step ``custom_prompt`` section (040): appended verbatim, no
+    rewording, under a fixed heading."""
+    return f"## {CUSTOM_PROMPT_HEADING}\n\n{text}"
 
 
 class _TurnPauseController:
@@ -432,15 +439,19 @@ def _artifact_file(runtime: Runtime, step_name: str) -> str | None:
 
 def build_context(runtime: Runtime, docs: dict[str, str], subject: str,
                   notes: list[str] | None = None,
-                  artifacts: dict[str, str] | None = None) -> str:
+                  artifacts: dict[str, str] | None = None,
+                  custom_prompt: str | None = None) -> str:
     """Assemble the attempt's shared message: the always-on agent-facing error
-    guide (034), subject, doc bundle, the canonical dependency-policy paragraph
-    (021), the expected-output-paths section generated from step defs (a prompt
-    merely naming files lets agents skip them), accumulated stage artifacts
-    (027), and previous-attempt notes."""
+    guide (034), subject, doc bundle, the run-specific instructions appended
+    from the step's ``custom_prompt`` (040), the canonical dependency-policy
+    paragraph (021), the expected-output-paths section generated from step
+    defs (a prompt merely naming files lets agents skip them), accumulated
+    stage artifacts (027), and previous-attempt notes."""
     parts = [f"You are working on {subject}.\n", guide_section()]
     for name, content in docs.items():
         parts.append(f"## {name}\n\n{content}")
+    if custom_prompt:
+        parts.append(custom_prompt_section(custom_prompt))
     if artifacts:
         parts.append("## Stage artifacts")
         for step_name, content in artifacts.items():
@@ -613,11 +624,14 @@ def build_brief_context(runtime: Runtime, step: StepConfig,
     """Assemble the brief agent's task message (033): the always-on
     agent-facing error guide (034), the resolved unit/plan docs (re-read from
     disk where sources are known), the closed ``brief_inputs.include`` file
-    set, and the env-facts block. Prior verdicts and feedback are deliberately
+    set, the run-specific instructions appended from ``custom_prompt`` (040),
+    and the env-facts block. Prior verdicts and feedback are deliberately
     absent - the notes channel delivers them per attempt."""
     sections, env_facts = _brief_sections(runtime, step, docs, doc_sources)
     parts = [guide_section()]
     parts += [f"## {name}\n\n{content}" for name, content in sections.items()]
+    if step.custom_prompt:
+        parts.append(custom_prompt_section(step.custom_prompt))
     if env_facts:
         parts.append("## Environment facts\n\n"
                      + "\n".join(f"- {fact}" for fact in env_facts))
@@ -683,7 +697,8 @@ def make_agent_node(runtime: Runtime, step: StepConfig):
         max_iterations = step.max_iterations or DEFAULT_MAX_ITERATIONS
         message = build_context(runtime, state["docs"], state["subject"],
                                 state["notes"],
-                                artifacts=state.get("artifacts", {}))
+                                artifacts=state.get("artifacts", {}),
+                                custom_prompt=step.custom_prompt)
         print(f"[{subject}] Running {step.agent} agent...", flush=True)
         result = runtime.run_agent(step.agent, message,
                                     max_iterations=max_iterations, step=step)

{agentic_dev_loop-1.0.0a37 → agentic_dev_loop-1.0.0a38/src/agentic_dev_loop.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agentic-dev-loop
-Version: 1.0.0a37
+Version: 1.0.0a38
 Summary: Turn an LLM into a gated, sandboxed, resumable construction loop over your codebase.
 Author: Luke Wilson
 License-Expression: Apache-2.0

{agentic_dev_loop-1.0.0a37 → agentic_dev_loop-1.0.0a38}/tests/test_033_brief_step.py

@@ -266,6 +266,18 @@ class TestBuildBriefContext:
         assert ctx.count("## AGENTS.md") == 1
         assert "REPO RULES FROM DOCS" in ctx
 
+    def test_custom_prompt_between_docs_and_env_facts(self, runtime):
+        step = make_cfg([{
+            "name": "brief", "agent": "brief", "produces": "attempt-brief.md",
+            "brief_inputs": {"include": ["AGENTS.md"], "env_facts": ["offline"]},
+            "custom_prompt": "brief scope", "on_pass": "success",
+        }]).steps[0]
+        ctx = build_brief_context(runtime, step, {"plan.md": "THE PLAN"})
+        assert "## Run-specific instructions\n\nbrief scope" in ctx
+        assert ctx.count("## Run-specific instructions") == 1
+        assert ctx.index("## plan.md") < ctx.index("## Run-specific instructions")
+        assert ctx.index("## Run-specific instructions") < ctx.index("## Environment facts")
+
 
 class TestBriefInputsHash:
     def test_stable_and_content_sensitive(self, runtime):

{agentic_dev_loop-1.0.0a37 → agentic_dev_loop-1.0.0a38}/tests/test_config.py

@@ -331,6 +331,66 @@ def test_timeout_without_command_rejected(tmp_path):
         _load(tmp_path, data)
 
 
+class TestCustomPrompt:
+    def test_defaults_none(self, tmp_path):
+        assert _load(tmp_path, _base()).steps[0].custom_prompt is None
+
+    def test_literal_text_round_trips(self, tmp_path):
+        data = _base()
+        data["steps"][0]["custom_prompt"] = "verify the export tab first"
+        assert _load(tmp_path, data).steps[0].custom_prompt == \
+            "verify the export tab first"
+
+    def test_existing_file_path_reads_contents(self, tmp_path):
+        (tmp_path / "qa-plan.md").write_text("verify the export tab\n")
+        data = _base()
+        data["steps"][0]["custom_prompt"] = "qa-plan.md"
+        assert _load(tmp_path, data).steps[0].custom_prompt == \
+            "verify the export tab\n"
+
+    def test_absolute_file_path_reads_contents(self, tmp_path):
+        plan = tmp_path / "plan.txt"
+        plan.write_text("ABSOLUTE PLAN")
+        data = _base()
+        data["steps"][0]["custom_prompt"] = str(plan)
+        assert _load(tmp_path, data).steps[0].custom_prompt == "ABSOLUTE PLAN"
+
+    def test_missing_path_with_extension_rejected(self, tmp_path):
+        data = _base()
+        data["steps"][0]["custom_prompt"] = "qa-plan.md"
+        with pytest.raises(ValueError,
+                           match="'custom_prompt' file not found: qa-plan.md"):
+            _load(tmp_path, data)
+
+    def test_missing_path_with_separator_rejected(self, tmp_path):
+        data = _base()
+        data["steps"][0]["custom_prompt"] = "tests/qa-plan"
+        with pytest.raises(ValueError, match="tests/qa-plan"):
+            _load(tmp_path, data)
+
+    def test_requires_agent(self, tmp_path):
+        data = _base()
+        data["steps"] = [{"name": "x", "command": "true",
+                          "custom_prompt": "hi"}]
+        with pytest.raises(ValidationError,
+                           match="'custom_prompt' requires 'agent'"):
+            _load(tmp_path, data)
+
+    def test_rejected_on_action_step(self, tmp_path):
+        data = _base()
+        data["steps"] = [{"name": "x", "action": "commit",
+                          "custom_prompt": "hi"}]
+        with pytest.raises(ValidationError,
+                           match="'custom_prompt' requires 'agent'"):
+            _load(tmp_path, data)
+
+    def test_empty_value_rejected(self, tmp_path):
+        data = _base()
+        data["steps"][0]["custom_prompt"] = ""
+        with pytest.raises(ValidationError, match="non-empty"):
+            _load(tmp_path, data)
+
+
 @pytest.mark.parametrize("field,value", [
     ("artifact", "a.md"),
     ("verdicts", {"pass": "P", "fail": "F"}),

{agentic_dev_loop-1.0.0a37 → agentic_dev_loop-1.0.0a38}/tests/test_stages.py

@@ -137,6 +137,19 @@ class TestBuildContext:
         ctx = build_context(runtime, {}, "s", [])
         assert "## Stage artifacts" not in ctx
 
+    def test_custom_prompt_section_between_prompt_and_artifacts(self, runtime):
+        ctx = build_context(runtime, {"plan.md": "THE PLAN"}, "s", [],
+                            artifacts={"dev": "summary"},
+                            custom_prompt="verify export first")
+        assert "## Run-specific instructions\n\nverify export first" in ctx
+        assert ctx.count("## Run-specific instructions") == 1
+        assert ctx.index("## plan.md") < ctx.index("## Run-specific instructions")
+        assert ctx.index("## Run-specific instructions") < ctx.index("## Stage artifacts")
+
+    def test_no_custom_prompt_leaves_context_unchanged(self, runtime):
+        ctx = build_context(runtime, {"plan.md": "THE PLAN"}, "s", [])
+        assert "## Run-specific instructions" not in ctx
+
 
 class TestDepsPolicyContext:
     def test_locked_policy_injected(self, runtime):
@@ -210,6 +223,17 @@ class TestAgentNode:
         update = make_agent_node(runtime, runtime.cfg.steps[0])(make_state())
         assert update["artifacts"] == {"dev": "MARKER-DEV"}
 
+    def test_custom_prompt_reaches_agent_message(self, runtime, fake_agents):
+        cfg = make_cfg([{"name": "dev", "agent": "dev",
+                         "produces": ["summary.md"],
+                         "custom_prompt": "scope note", "on_pass": "success"}])
+        rt = dataclasses.replace(runtime, cfg=cfg)
+        fake_agents.set("dev", [("write", "summary.md", "done")])
+        make_agent_node(rt, rt.cfg.steps[0])(make_state())
+        (message,) = fake_agents.messages_for("dev")
+        assert "## Run-specific instructions\n\nscope note" in message
+        assert message.count("## Run-specific instructions") == 1
+
     def test_pass_verdict_captures_artifact(self, runtime, fake_agents):
         fake_agents.set("check", [("write", "ci-fix.md", "MARKER-CHECK\nVERDICT: PASS\n")])
         update = make_agent_node(runtime, runtime.cfg.steps[1])(make_state())