feat: Milestone 0 — project foundation and package skeleton (v0.1.0) by shaypal5 · Pull Request #3 · leadforge-dev/leadforge

shaypal5 · 2026-04-18T08:34:54Z

Summary

Bootstraps the leadforge codebase from empty to a fully installable, lint-clean, tested package skeleton. This is the complete Milestone 0 delivery.

All acceptance criteria pass:

pip install -e . ✓
leadforge --help shows all four commands ✓
leadforge list-recipes returns b2b_saas_procurement_v1 ✓
20 tests passing ✓
ruff + mypy clean ✓
CI configured ✓

What's included

Package scaffold

pyproject.toml — setuptools build, Typer + PyYAML runtime, ruff/mypy/pytest dev deps, leadforge CLI entry point
Full subpackage skeleton (api, cli, core, narrative, schema, structure, mechanisms, simulation, render, exposure, validation, recipes) with __init__.py stubs
README.md — install, CLI quickstart, Python API snippet, doc links
.pre-commit-config.yaml — ruff (lint + format) + pre-commit-hooks

Core primitives (leadforge/core/)

enums.py — ExposureMode + DifficultyProfile as StrEnum
exceptions.py — LeadforgeError base + 6 typed subclasses
models.py — GenerationConfig, WorldSpec, WorldBundle dataclass stubs
rng.py, ids.py — documented stubs (implemented in M1)

Recipe system (leadforge/recipes/)

registry.py — list_recipes() + load_recipe() reading YAML files
b2b_saas_procurement_v1/recipe.yaml — id, title, primary task, modes, difficulty levels

CLI (leadforge/cli/)

main.py — Typer app with --version and four registered commands
list-recipes — fully implemented with Rich table output
generate — stub with full option spec per architecture doc; exits 1 with "coming in v0.2.0"
inspect, validate — stubs with correct argument spec

CI (.github/workflows/ci.yml)

Three jobs: lint (ruff check + format), typecheck (mypy), test matrix (Python 3.11 + 3.12)
Coverage artifacts uploaded under pr-agent-context-coverage-py* prefix for pr-agent-context integration

Tests (20 passing)

tests/test_cli.py — help, version, list-recipes content, stub exit codes
tests/core/test_enums.py — values + string construction
tests/core/test_exceptions.py — hierarchy + message preservation
tests/recipes/test_registry.py — list/load, required fields, error case

Test plan

CI passes (lint, typecheck, test matrix)
leadforge --help shows all four commands
leadforge list-recipes shows b2b_saas_procurement_v1
leadforge --version prints leadforge 0.1.0

🤖 Generated with Claude Code

Bootstraps the leadforge codebase from empty to a fully installable, testable, lint-clean package skeleton. All Milestone 0 acceptance criteria pass: `pip install -e .` works, `leadforge --help` shows all four commands, `leadforge list-recipes` returns the v1 recipe, and CI is configured. Package scaffold - pyproject.toml: setuptools build, Typer+PyYAML runtime deps, ruff/mypy/ pytest dev deps, `leadforge` CLI entry point - Full subpackage skeleton with __init__.py stubs for every module in the canonical layout (api, cli, core, narrative, schema, structure, mechanisms, simulation, render, exposure, validation, recipes) - leadforge/version.py: __version__ = "0.1.0" - README.md: install, quickstart, API snippet, doc links - .pre-commit-config.yaml: ruff (lint+format) + pre-commit-hooks Core primitives (leadforge/core/) - enums.py: ExposureMode (StrEnum), DifficultyProfile (StrEnum) - exceptions.py: LeadforgeError base + 6 typed subclasses - models.py: GenerationConfig, WorldSpec, WorldBundle dataclass stubs - rng.py, ids.py: documented stubs for Milestone 1 Recipe system (leadforge/recipes/) - registry.py: list_recipes() + load_recipe() reading from YAML files - b2b_saas_procurement_v1/recipe.yaml: id, title, primary_task, supported_modes, supported_difficulty, default_population CLI (leadforge/cli/) - main.py: Typer app with --version flag and four registered commands - commands/list_recipes.py: fully implemented with Rich table output - commands/generate.py: stub with full option spec (--recipe, --seed, --mode, --out, --difficulty, --n-accounts, --n-contacts, --n-leads, --horizon-days, --override); exits 1 with "coming in v0.2.0" - commands/inspect.py, validate.py: stubs with correct argument spec CI (.github/workflows/ci.yml) - Three jobs: lint (ruff check+format), typecheck (mypy), test matrix (Python 3.11 + 3.12 with pytest-cov + coverage artifact upload for pr-agent-context integration) Tests (20 passing) - tests/test_cli.py: help, version, list-recipes output, stub exit codes - tests/core/test_enums.py: values and string construction - tests/core/test_exceptions.py: hierarchy and message preservation - tests/recipes/test_registry.py: list/load, required fields, error case Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Bootstraps the leadforge repository into an installable Python package with a Typer-based CLI, an initial YAML-backed recipe registry, CI/tooling configuration, and a baseline test suite to support future milestones.

Changes:

Added project packaging/tooling (pyproject.toml, pre-commit, CI) and a full package/module skeleton.
Implemented a first working CLI command (list-recipes) plus stubs for generate, inspect, validate.
Added a YAML recipe registry + an initial b2b_saas_procurement_v1 recipe and accompanying tests.

Reviewed changes

Copilot reviewed 23 out of 44 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tests/test_cli.py	CLI smoke tests for help/version and command stubs
tests/recipes/test_registry.py	Tests for recipe listing/loading and error case
tests/recipes/init.py	Test package marker
tests/core/test_exceptions.py	Exception hierarchy tests
tests/core/test_enums.py	Enum value/from-string tests
tests/core/init.py	Test package marker
tests/init.py	Test package marker
pyproject.toml	Packaging metadata, deps, ruff/mypy/pytest configuration
leadforge/version.py	Defines `__version__`
leadforge/validation/init.py	Subpackage stub
leadforge/structure/init.py	Subpackage stub
leadforge/simulation/init.py	Subpackage stub
leadforge/schema/init.py	Subpackage stub
leadforge/sample_data/public/.gitkeep	Placeholder for sample data
leadforge/sample_data/instructor/.gitkeep	Placeholder for sample data
leadforge/render/init.py	Subpackage stub
leadforge/recipes/registry.py	Implements recipe discovery/loading from YAML
leadforge/recipes/b2b_saas_procurement_v1/recipe.yaml	Adds initial recipe metadata YAML
leadforge/recipes/b2b_saas_procurement_v1/init.py	Subpackage stub
leadforge/recipes/init.py	Subpackage stub
leadforge/narrative/init.py	Subpackage stub
leadforge/mechanisms/init.py	Subpackage stub
leadforge/exposure/init.py	Subpackage stub
leadforge/examples/notebooks/.gitkeep	Placeholder for examples
leadforge/examples/configs/.gitkeep	Placeholder for examples
leadforge/core/rng.py	RNG utilities stub docstring
leadforge/core/models.py	Dataclass stubs for config/spec/bundle
leadforge/core/ids.py	ID scheme stub docstring
leadforge/core/exceptions.py	Defines project exception hierarchy
leadforge/core/enums.py	Defines `ExposureMode` / `DifficultyProfile`
leadforge/core/init.py	Core subpackage stub
leadforge/cli/main.py	Typer app entrypoint + command registration
leadforge/cli/commands/validate.py	`validate` command stub
leadforge/cli/commands/list_recipes.py	`list-recipes` command implementation (Rich table)
leadforge/cli/commands/inspect.py	`inspect` command stub
leadforge/cli/commands/generate.py	`generate` command stub + option surface
leadforge/cli/commands/init.py	Commands subpackage stub
leadforge/cli/init.py	CLI subpackage stub
leadforge/api/init.py	API subpackage stub
leadforge/init.py	Package init exporting `__version__`
README.md	Project README with install/quickstart/docs links
.pre-commit-config.yaml	Pre-commit hooks (ruff, formatting, basic checks)
.github/workflows/ci.yml	CI for lint/typecheck/tests with coverage artifacts
.agent-plan.md	Updates project plan to mark M0 complete / M1 next

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

COPILOT-1 — Add Generator stub to leadforge/api - leadforge/api/generator.py: Generator class with from_recipe() and generate() raising NotImplementedError with "coming in v0.2.0" messages - leadforge/api/__init__.py: export Generator so `from leadforge.api import Generator` resolves correctly COPILOT-2 — Sort list_recipes() by recipe id field, not path - return sorted(recipes, key=lambda r: r["id"]) instead of relying on filesystem iteration order COPILOT-3 — Validate yaml.safe_load() result in registry - Extract _parse_and_validate() helper; raises InvalidRecipeError if the parsed value is not a dict or is missing the required 'id' key; used by both list_recipes() and load_recipe() COPILOT-4 — Guard load_recipe() against path traversal - Resolve the candidate path and verify it stays within _RECIPES_DIR before checking existence or opening; raises InvalidRecipeError for any recipe_id that would escape the recipes directory COPILOT-5 — Comment out unimplemented CLI commands in README - Quickstart now shows generate/inspect/validate as commented-out examples with "Coming in v0.x.0" labels; only `list-recipes` is shown as immediately runnable - Python API snippet annotated with "(coming in v0.2.0)" Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 25 out of 45 changed files in this pull request and generated 2 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

…mits Documents the mandatory step of resolving GitHub review threads via GraphQL after addressing PR comments, so the omission from PR #3 does not recur. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- registry.py: replace string-prefix path traversal guard with Path.is_relative_to() (Python 3.11+), closing the prefix-collision bypass (e.g. recipes_evil alongside recipes) - pyproject.toml: add "S" (bandit) ruleset to ruff select so security checks are active on non-test code; widen per-file-ignores glob from tests/* to tests/**/* to cover subdirectories; add S108 to test ignores to suppress the /tmp false-positive in test CLI invocations Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

github-actions · 2026-04-18T14:08:32Z

pr-agent-context report:

This run includes unresolved review comments on PR #3.

For each unresolved review comment, recommend one of: resolve as irrelevant, accept and implement
the recommended solution, open a separate issue and resolve as out-of-scope for this PR, accept and
implement a different solution, or resolve as already treated by the code.

After I reply with my decision per item, implement the accepted actions, resolve the corresponding
PR comments, and push all of these changes in a single commit.

# Copilot Comments

## COPILOT-1
Location: leadforge/recipes/registry.py
URL: https://github.com/leadforge-dev/leadforge/pull/3#discussion_r3105050071
Status: outdated
Root author: copilot-pull-request-reviewer

Comment:
    The path traversal guard is bypassable because it uses a string prefix check. For example, a resolved path like `/.../recipes_evil` will still start with `/.../recipes`, so `recipe_id='../recipes_evil'` can incorrectly pass this check. Use `Path.is_relative_to()` (Py3.11+) or `recipe_dir.relative_to(base_dir)` in a try/except to ensure the resolved path is actually within `_RECIPES_DIR`.
    ~~~suggestion
        base_dir = _RECIPES_DIR.resolve()
        recipe_dir = (base_dir / recipe_id).resolve()
        try:
            recipe_dir.relative_to(base_dir)
        except ValueError:
    ~~~

## COPILOT-2
Location: pyproject.toml
URL: https://github.com/leadforge-dev/leadforge/pull/3#discussion_r3105050075
Status: outdated
Root author: copilot-pull-request-reviewer

Comment:
    The Ruff per-file-ignores glob `tests/*` only matches files directly under `tests/` and won’t apply to tests in subdirectories like `tests/core/...` or `tests/recipes/...`. If the intent is to ignore rules across the whole test tree, use `tests/**` (or similar). Also, `S101` is currently a no-op unless the `S` ruleset is enabled in `select`.

Run metadata:

Tool ref: v4
Tool version: 4.0.18
Trigger: commit pushed
Workflow run: 24606387604 attempt 1
Comment timestamp: 2026-04-18T14:07:41.401806+00:00
PR head commit: ac22ad52c0c25ff581cb8af14d2a97de9b664c63

Copilot

Pull request overview

Copilot reviewed 26 out of 46 changed files in this pull request and generated 3 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Review feedback addressed: - Remove primary_task/label_window_days as explicit kwargs from resolve_config() and Generator.from_recipe() — these fields are resolved from recipe YAML and override dict only, not casually overridable, since the generation pipeline doesn't yet support arbitrary task types (Copilot-1, Copilot-3, shaypal5 #1, #2) - Add label_window_days <= horizon_days validation in GenerationConfig.__post_init__ (Copilot-2, shaypal5 #3) - Add tests for invalid primary_task on GenerationConfig: empty string, non-string type (shaypal5 #6, pr-agent-context) - Add tests for invalid label_window_days on Recipe.from_dict: bool, non-positive, float (shaypal5 #7, pr-agent-context) - Add test for label_window_days > horizon_days rejection - Fix existing test using horizon_days=30 (now conflicts with default label_window_days=90) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* feat: carry primary_task and label_window_days into WorldSpec for dataset card Add `primary_task` and `label_window_days` fields to `GenerationConfig` (with defaults preserving current behavior). Propagate through `Recipe.from_dict()`, `resolve_config()`, and `Generator.from_recipe()` so recipe YAML can override them. Update `render_dataset_card()` to read from `world_spec.config` instead of hard-coded string literals. Closes #6 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * docs: update .agent-plan.md for WorldSpec task fields (PR #36) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: address review feedback — tighten scope, add validation + tests Review feedback addressed: - Remove primary_task/label_window_days as explicit kwargs from resolve_config() and Generator.from_recipe() — these fields are resolved from recipe YAML and override dict only, not casually overridable, since the generation pipeline doesn't yet support arbitrary task types (Copilot-1, Copilot-3, shaypal5 #1, #2) - Add label_window_days <= horizon_days validation in GenerationConfig.__post_init__ (Copilot-2, shaypal5 #3) - Add tests for invalid primary_task on GenerationConfig: empty string, non-string type (shaypal5 #6, pr-agent-context) - Add tests for invalid label_window_days on Recipe.from_dict: bool, non-positive, float (shaypal5 #7, pr-agent-context) - Add test for label_window_days > horizon_days rejection - Fix existing test using horizon_days=30 (now conflicts with default label_window_days=90) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Fold the brutal self-review's findings back into the PR before review. Bugs: - (#1) run_packager validate→write order — both packagers wrote README/metadata on validation failure, leaving corrupt artifacts on disk that would silently get committed. Gated on `errors == ()`; added no-write tests for both packagers. - (#2) Instructor README inlined the public 3-tier README into a 1-tier dataset card. Replaced with a dedicated `INSTRUCTOR_BODY` constant that links to the public dataset and describes only the instructor-specific additions (full-horizon tables, hidden DAG, latent registry, mechanism summary). - (#3) validate_upload_dir_safe also blocks strict descendants of release_dir; `--huggingface-dir release/intro` would otherwise rmtree the intro bundle. Architecture: - (#5) Finished shared-primitives extraction: SOURCE_TREE_BLOCK, validate_readme_substitution, replace_file, replace_dir, load_manifest now live in scripts/_release_common.py. Both packagers reduced to imports. - (#6) Replaced 60-line hand-rolled YAML renderer with yaml.safe_dump + a 4-line _IndentedDumper subclass. - (#7) Removed dead --owner / --dataset-slug CLI flags. - (#8) assemble_upload_dir now takes rendered_readme and writes it. - (#9) build_config_for_tier made pure (no I/O); cheap manifest-stat preflight via _assert_tier_dir_exists. - (#10) --default-config with --variant=instructor errors loudly. CI: - (#4) Added [publish] extra (datasets>=2.14, kaggle>=1.6) so the gated G12.3 / G12.4 / G11.3 tests install in one line. Cleanups: visual cruft (#13–#16), test cruft (#17 — unused tmp_path, dead tag_lines), em-dash YAML round-trip parametrised for the instructor pretty_name. Verification: 1223 tests pass + 5 gated skips; ruff + mypy clean; hash determinism PASS 67/67; leakage probes 0/3 reconstruct on every tier; validate_release_candidate --no-rebuild exits 0. release/{kaggle,huggingface,huggingface-instructor}/dataset-metadata .json|README.md regenerated; audit-artifact-sync tests guard them. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* PR 5.2: HuggingFace release packager + load_dataset smoke test Add `scripts/package_hf_release.py` to generate `release/huggingface/README.md` with G12.1-compliant YAML frontmatter (pretty_name, license, language, task_categories, size_categories, tags, three configs with `default: true` on intermediate per G12.2), inlining the rewritten `release/README.md` body with HF-specific link rewrites. `--variant=instructor` packages the companion repo (G12.4) from `release/intermediate_instructor/` into a separate `release/huggingface-instructor/` upload tree. G12.3 covered by a parametrised `load_dataset()` smoke test gated on the optional `datasets` SDK. Extract shared release-packaging primitives (link rewriter, dir-safety guard, cover-image validator) into `scripts/_release_common.py`; refactor the Kaggle packager to import them. `release/kaggle/dataset-metadata.json` is byte-stable across the refactor. Delete the legacy `release/HF_DATASET_CARD.md` stub — superseded by the generated card. Gitignore `release/huggingface{,-instructor}/*` except the committed README. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * PR 5.2 self-review fixes (Kaggle + HF packagers) Fold the brutal self-review's findings back into the PR before review. Bugs: - (#1) run_packager validate→write order — both packagers wrote README/metadata on validation failure, leaving corrupt artifacts on disk that would silently get committed. Gated on `errors == ()`; added no-write tests for both packagers. - (#2) Instructor README inlined the public 3-tier README into a 1-tier dataset card. Replaced with a dedicated `INSTRUCTOR_BODY` constant that links to the public dataset and describes only the instructor-specific additions (full-horizon tables, hidden DAG, latent registry, mechanism summary). - (#3) validate_upload_dir_safe also blocks strict descendants of release_dir; `--huggingface-dir release/intro` would otherwise rmtree the intro bundle. Architecture: - (#5) Finished shared-primitives extraction: SOURCE_TREE_BLOCK, validate_readme_substitution, replace_file, replace_dir, load_manifest now live in scripts/_release_common.py. Both packagers reduced to imports. - (#6) Replaced 60-line hand-rolled YAML renderer with yaml.safe_dump + a 4-line _IndentedDumper subclass. - (#7) Removed dead --owner / --dataset-slug CLI flags. - (#8) assemble_upload_dir now takes rendered_readme and writes it. - (#9) build_config_for_tier made pure (no I/O); cheap manifest-stat preflight via _assert_tier_dir_exists. - (#10) --default-config with --variant=instructor errors loudly. CI: - (#4) Added [publish] extra (datasets>=2.14, kaggle>=1.6) so the gated G12.3 / G12.4 / G11.3 tests install in one line. Cleanups: visual cruft (#13–#16), test cruft (#17 — unused tmp_path, dead tag_lines), em-dash YAML round-trip parametrised for the instructor pretty_name. Verification: 1223 tests pass + 5 gated skips; ruff + mypy clean; hash determinism PASS 67/67; leakage probes 0/3 reconstruct on every tier; validate_release_candidate --no-rebuild exits 0. release/{kaggle,huggingface,huggingface-instructor}/dataset-metadata .json|README.md regenerated; audit-artifact-sync tests guard them. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * PR 5.2 Copilot-review fixes (Kaggle + HF packagers) Fold Copilot's two real findings on the self-review revision back in. COPILOT-1 — validate_upload_dir_safe was only invoked inside assemble_upload_dir, which --dry-run skips. A dry-run with --huggingface-dir release (or .) would write the README into the unsafe path BEFORE the safety net fired. Hoist the check into run_packager (both packagers) so it runs before any mkdir or write; the inner assemble_upload_dir call stays as defence-in-depth for direct callers. New tests: dry-run with unsafe upload-dir raises without writing; the same path through main() returns rc=2. COPILOT-2 — Cover-image path resolution was inconsistent: validate_cover_image used cover_image as passed, while assemble_upload_dir did a separate ``release_dir / cover_image.name`` fallback. Diverged for bare-basename inputs (false validation failures) and two-paths-sharing-a-basename (assembler shadowing the explicit path). Added resolve_cover_image_path() to _release_common.py (explicit-wins, release-dir fallback); run_packager calls it once and threads the resolved path through validation, the metadata's image field, and assembly. New tests/scripts/test_release_common.py covers the four resolution branches; new packager-side tests confirm bare-basename success + metadata field plumbing. COPILOT-3 — outdated; already addressed by self-review fix #8 in commit f2fc4a2. Resolved as already treated; no code change. Verification: 1232/1232 tests pass + 5 gated skips; ruff + mypy clean; hash determinism PASS 67/67; leakage probes rc=0 on every tier; validate_release_candidate --no-rebuild exits 0; BUNDLE_SCHEMA_VERSION unchanged at 5. release/{kaggle,huggingface,huggingface-instructor}/* artifacts regenerated byte-identically. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

…1] (#121) * refactor(render): scheme-agnostic build_manifest + schema v6 [LTV-Pn.1] First sub-PR of the split LTV-Pn (M6). Decouples the manifest builder from the lead-scoring scheme so the lifecycle scheme can reuse it, and records which scheme produced each bundle. - build_manifest no longer takes the lead-scoring `world_graph`. It takes `generation_scheme: str` (required), `motif_family: str | None = None`, and an `extra_fields` mapping for scheme-specific top-level keys (the lifecycle scheme will add `observation_date` / forward windows). A collision guard rejects extra_fields that would clobber a core key. - Every manifest now records `generation_scheme` (lead_scoring / lifecycle). - BUNDLE_SCHEMA_VERSION 5 -> 6 (history note added). The lead-scoring published *shape* is unchanged: tables/ and tasks/ parquet files are byte-identical to v5 (verified via pinned-timestamp SHA-256 across the full instructor bundle); only manifest.json changes (new field + version). - Removes the render.manifests -> lead_scoring.structure.graph TYPE_CHECKING back-reference (partial discharge of carried cleanup #3; the core.models.WorldBundle back-refs follow in Pn.2). - Schema-contract gate renamed test_bundle_schema_v5_contract.py -> _v6_, asserts version "6", and adds a generation_scheme assertion. The pinned column/table sets carry over verbatim (shape unchanged). Callers updated: lead_scoring.write_bundle passes generation_scheme=self.name, motif_family=world_graph.motif_family. Plan: docs/ltv/roadmap.md records the four-way LTV-Pn split (Pn.1…4) and marks the cleanup-#3 partial discharge. Tests: new tests/render/test_manifest_scheme_agnostic.py (generation_scheme recorded, motif_family default/passthrough, extra_fields merge + collision guard, required-arg); v6 contract test. Full suite 1801 passed / 51 skipped; ruff + mypy clean. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix(cli): surface generation_scheme in inspect; honest fixtures [LTV-Pn.1] Self-review findings on the manifest-generalization PR. 1. Verification gap (mine): the byte-identity proof covered only the research_instructor bundle. Confirmed the student_public path — which takes the snapshot-safe relational route and also calls build_manifest — is unchanged and deterministic, and carries generation_scheme + v6. No code change. 2. The inspect CLI is the user-facing manifest viewer but ignored the new first-class generation_scheme field. Added a conditional "Scheme:" row (conditional so pre-v6 bundles still render without a "?" placeholder, matching the existing v3+/v4+ field handling). Also made the motif_family row null-safe: a lifecycle bundle records motif_family=null, which .get(..., '?') would render as the literal "None"; now shows "?". 3. Bumped four release/critique test fixtures from a hardcoded bundle_schema_version "5" to "6". They are inert (nothing compares them to the current version, which is why the suite stayed green), but a fixture claiming "5" while the package emits "6" is misleading. Tests: inspect-shows-generation-scheme CLI test; the header-order regression guard still holds (the new row sits between Recipe and Seed without reordering the pinned eight). Full suite 1802 passed / 51 skipped; ruff + mypy clean. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * docs(test): fix stale v5 docstring in snapshot-safe integration test [LTV-Pn.1] Copilot review on #121: the module docstring opening line still said "bundle schema v5" while line 9 (updated in this PR) references v6. Make the version consistent. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

…n.2] (#122) * refactor: scheme-agnostic WorldBundle + exposure metadata hook [LTV-Pn.2] Second sub-PR of the split LTV-Pn. Removes the last core/shared-layer references to the lead-scoring scheme (carried cleanups #2 and #3), so the lifecycle scheme can plug into the same envelope. WorldBundle (cleanup #3): - Replaced the three lead-scoring-typed fields (population / simulation_result / world_graph) with a single opaque `artifacts: Any`. Each scheme defines and unwraps its own container; lead-scoring adds LeadScoringArtifacts (schemes/lead_scoring/artifacts.py). core.models no longer imports any lead_scoring type — the layering inversion introduced in LTV-Pf.1 is gone. Exposure (cleanup #2): - apply_exposure is now scheme-agnostic: it writes the generic, spec-only world_spec.json (kept in exposure/metadata.py as write_world_spec_json) and dispatches the scheme-specific hidden-truth files to a new GenerationScheme.write_metadata(bundle, meta_dir) hook, resolved from bundle.spec.scheme via the registry. The lead-scoring graph / latent registry / mechanism-summary writers moved out of exposure/ into LeadScoringScheme.write_metadata. exposure/ no longer references lead_scoring. - Protocol gains write_metadata; the lifecycle stub implements it (raises NotImplementedError until Pn.4). Byte-identity: the full lead-scoring bundle is byte-identical across BOTH exposure modes (research_instructor 21 files, student_public 14) — verified against a pre-refactor SHA-256 reference. The metadata writers moved, not changed; world_spec.json content/order is preserved. Re-scope: the shared bundle orchestrator (cleanup #1) moves from this PR to LTV-Pn.4. Per the roadmap's own note it is best designed with the second scheme's write_bundle in hand; extracting it now against one scheme would guess the hook shape. Roadmap + deferred-cleanups updated (#2, #3 done; #1 → Pn.4). Tests: new tests/schemes/test_scheme_metadata_hook.py (artifacts populated; WorldBundle has only spec+artifacts; generic world_spec writer; lead-scoring hook emits the 4 hidden-truth files; unpopulated-bundle + lifecycle-stub raise). Updated field-access sites in 5 test modules. Full suite 1808 passed / 51 skipped; ruff + mypy clean. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix(scripts): migrate build_v*_snapshot to bundle.artifacts [LTV-Pn.2] Self-review finding on the WorldBundle generalization: five dataset-build scripts still read the removed bundle.simulation_result / bundle.population fields and would AttributeError at runtime. This slipped both nets: - mypy leadforge does not type-check scripts/, and - the existing tests/scripts/test_build_v*_snapshot.py cover only the pipelines.build_v* transform helpers (pure DataFrame functions), never the generate_bundle() entry point that touches the bundle. CI's "Validate v6/v7" jobs validate a pre-built CSV; they do not regenerate, so they would not have caught it either. Fix: build_v4/v5/v6/v7_snapshot.py and build_midproject_lead_scoring.py now read bundle.artifacts.simulation_result / bundle.artifacts.population. Smoke- ran the v6 builder end-to-end through Generator.generate() to confirm. Guard: tests/scripts/test_build_v6_snapshot.py gains TestGenerateBundleArtifactsPath, which loads the script via importlib and runs generate_bundle(small) so a future WorldBundle field rename can't silently break the generate path again. The other four scripts share the identical access pattern. Full suite 1809 passed / 51 skipped; ruff + mypy clean. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * fix(exposure): clear metadata/ before rewrite; fix stale docstring [LTV-Pn.2] Two Copilot review findings on #122. 1. apply_exposure reused an existing metadata/ via mkdir(exist_ok=True) when writing hidden truth, so a reused output path could retain stale files. Pre-existing behavior, but Pn.2 makes it newly dangerous: once the lifecycle scheme writes a different hidden-truth file set, regenerating a lifecycle bundle over a path that previously held a lead-scoring bundle would orphan graph.graphml / mechanism_summary.json into the new bundle. Now apply_exposure always removes any existing metadata/ first, then recreates it when writing — so contents exactly match the current bundle (mirroring the non-writing branch, which already rmtree'd it). Byte-identity preserved for both modes (fresh paths have no metadata/, so the rmtree is a guarded no-op). Regression tests: a pre-seeded stale file is gone after an instructor rewrite; student_public still removes the dir entirely. 2. The lead_scoring.write_bundle docstring still described apply_exposure as writing the lead-scoring hidden graph + latent registry directly. Updated to reflect the Pn.1/Pn.2 reality: build_manifest and apply_exposure are scheme-agnostic, and hidden truth is delegated to write_metadata; the remaining shared-orchestrator extraction is deferred to Pn.4. Full suite 1811 passed / 51 skipped; ruff + mypy clean. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

shaypal5 added this to the v0.1.0 — Repo & CLI skeleton milestone Apr 18, 2026

shaypal5 added type: feature New capability layer: core core/ primitives (RNG, IDs, models, exceptions) layer: cli cli/ command-line interface layer: recipes recipes/ recipe assets and registry labels Apr 18, 2026

Copilot AI review requested due to automatic review settings April 18, 2026 08:34

shaypal5 added type: feature New capability layer: core core/ primitives (RNG, IDs, models, exceptions) layer: cli cli/ command-line interface layer: recipes recipes/ recipe assets and registry labels Apr 18, 2026

Copilot started reviewing on behalf of shaypal5 April 18, 2026 08:35 View session

This comment has been minimized.

Sign in to view

Copilot AI reviewed Apr 18, 2026

View reviewed changes

Comment thread README.md

Comment thread leadforge/recipes/registry.py Outdated

Comment thread leadforge/recipes/registry.py Outdated

Comment thread leadforge/recipes/registry.py Outdated

Comment thread README.md Outdated

This comment has been minimized.

Sign in to view

shaypal5 self-assigned this Apr 18, 2026

This comment has been minimized.

Sign in to view

shaypal5 requested a review from Copilot April 18, 2026 11:09

Copilot started reviewing on behalf of shaypal5 April 18, 2026 11:09 View session

Copilot AI reviewed Apr 18, 2026

View reviewed changes

Comment thread leadforge/recipes/registry.py Outdated

Comment thread pyproject.toml Outdated

This comment has been minimized.

Sign in to view

Copilot AI review requested due to automatic review settings April 18, 2026 14:07

Copilot started reviewing on behalf of shaypal5 April 18, 2026 14:07 View session

shaypal5 merged commit da98878 into main Apr 18, 2026
9 checks passed

shaypal5 deleted the feat/milestone-0-foundation branch April 18, 2026 14:10

Copilot AI reviewed Apr 18, 2026

View reviewed changes

Comment thread leadforge/recipes/registry.py

Comment thread tests/test_cli.py

Comment thread tests/test_cli.py

shaypal5 mentioned this pull request May 4, 2026

Structural leakage in student_public bundles (post-#56 follow-up) #57

Closed

shaypal5 mentioned this pull request May 6, 2026

PR 5.2: HuggingFace release packager + load_dataset smoke test #72

Merged

4 tasks

shaypal5 mentioned this pull request Jun 12, 2026

refactor(render): scheme-agnostic build_manifest + schema v6 [LTV-Pn.1] #121

Merged

shaypal5 mentioned this pull request Jun 12, 2026

refactor: scheme-agnostic WorldBundle + exposure metadata hook [LTV-Pn.2] #122

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: Milestone 0 — project foundation and package skeleton (v0.1.0)#3

feat: Milestone 0 — project foundation and package skeleton (v0.1.0)#3
shaypal5 merged 4 commits into
mainfrom
feat/milestone-0-foundation

shaypal5 commented Apr 18, 2026

Uh oh!

This comment has been minimized.

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

This comment has been minimized.

github-actions Bot commented Apr 18, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

shaypal5 commented Apr 18, 2026

Summary

What's included

Test plan

Uh oh!

This comment has been minimized.

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

This comment has been minimized.

github-actions Bot commented Apr 18, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants