Make 1.0 a runner-only framework for coding-agent users

.github/workflows/dependency-review.yml

@@ -17,7 +17,7 @@ jobs:
         uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
 
       - name: Dependency Review
-        uses: actions/dependency-review-action@a1d282b36b6f3519aa1f3fc636f609c47dddb294 # v4
+        uses: actions/dependency-review-action@a1d282b36b6f3519aa1f3fc636f609c47dddb294 # v5.0.0
         with:
           fail-on-severity: high
           comment-summary-in-pr: always

.github/workflows/docs.yml

@@ -20,13 +20,13 @@ on:
   workflow_call:
     inputs:
       version:
-        description: "Version tag (e.g. v0.6.23) to validate and build from"
+        description: "Version tag (e.g. v1.0.0) to validate and build from"
         required: false
         type: string
   workflow_dispatch:
     inputs:
       version:
-        description: "Version tag (e.g. v0.6.23) to validate and build from"
+        description: "Version tag (e.g. v1.0.0) to validate and build from"
         required: false
         type: string
 

.gitignore

@@ -1,67 +1,65 @@
-# Python
-__pycache__/
-*.py[cod]
-*$py.class
-*.so
-.Python
-build/
-develop-eggs/
-dist/
-downloads/
-eggs/
-.eggs/
-lib/
-lib64/
-parts/
-sdist/
-var/
-wheels/
-*.egg-info/
-.installed.cfg
-*.egg
-
-# Virtual environments
-.venv/
-venv/
-ENV/
-env/
-
-# IDE
-.idea/
-.vscode/
-*.swp
-*.swo
-*~
-
-# Testing
-.pytest_cache/
-.coverage
-htmlcov/
-.tox/
-.nox/
-
-# Type checking
-.mypy_cache/
-
-# Test outputs
-test_results/
-aitest-reports/
-tests/fixtures/reports/html/
-
-# Node.js (Tailwind build)
-node_modules/
-src/pytest_aitest/templates/node_modules/
-package-lock.json
-
-# Environment
-.env
-.env.local
-
-# OS
-.DS_Store
-Thumbs.db
-site/
-.cache/
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# Virtual environments
+.venv/
+venv/
+ENV/
+env/
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+*~
+
+# Testing
+.pytest_cache/
+.coverage
+htmlcov/
+.tox/
+.nox/
+
+# Type checking
+.mypy_cache/
+
+# Test outputs
+test_results/
+aitest-reports/
+
+# Node.js
+node_modules/
+package-lock.json
+
+# Environment
+.env
+.env.local
+
+# OS
+.DS_Store
+Thumbs.db
+site/
+.cache/
 # Squad: ignore runtime state (logs, inbox, sessions)
 .squad/orchestration-log/
 .squad/log/
@@ -71,4 +69,4 @@ site/
 .squad-workstream
 
 # pytest-skill-engineering test scratch (some server tests create a cwd dir)
-.pytest-execution-servers/
+.pytest-execution-servers/

.squad/agents/fenster/charter.md

@@ -12,9 +12,9 @@
 ## What I Own
 
 - `src/pytest_skill_engineering/core/` — Skill, errors, and core types
-- `src/pytest_skill_engineering/copilot/` — CopilotEval, runner, fixtures, judge, result
+- `src/pytest_skill_engineering/copilot/` — CopilotEval, runner, controls, fixtures, result
 - `src/pytest_skill_engineering/execution/` — MCPServer, CLIServer, MCPServerProcess, CLIServerProcess
-- `src/pytest_skill_engineering/fixtures/` — factories, llm_assert, llm_score, skill fixtures
+- `src/pytest_skill_engineering/fixtures/` — factories and repetition support
 - Plugin mechanics (`plugin.py`, markers, hooks)
 
 ## How I Work
@@ -29,7 +29,7 @@
 
 **I handle:** Engine internals, core types, execution pipeline, MCP/CLI server management, Copilot SDK integration, plugin mechanics.
 
-**I don't handle:** HTML reports (McManus), integration tests (Hockney), architecture decisions without Keaton's input.
+**I don't handle:** Native evidence serialization (McManus), integration tests (Hockney), architecture decisions without Keaton's input.
 
 **When I'm unsure:** I say so and suggest who might know.
 

.squad/agents/hockney/charter.md

@@ -12,24 +12,25 @@
 ## What I Own
 
 - `tests/integration/copilot/` — All CopilotEval integration tests
-- `tests/showcase/` — Hero report tests
+- `examples/skill-dogfood/` — Customer workflow and historical skill-removal case study
+- `tests/contracts/` — Offline framework boundaries, not model performance
 - `tests/integration/conftest.py` — Shared constants and server fixtures
 - Test harnesses in `src/pytest_skill_engineering/testing/` — banking, todo MCP servers
 
 ## How I Work
 
-- **Integration tests only** — unit tests with mocked LLMs are worthless for this project
+- Real integration tests establish AI-interface behavior; offline contracts establish framework controls and evidence fidelity, not model performance
 - Always use `uv run python -m pytest` — bare `pytest` won't find the installed package
 - Run test files ONE AT A TIME, sequentially: start with `test_01_basic.py`, fix all, move to next
 - Use `--lf` to re-run only failures after a full run
-- Fast execution (< 1 second) is a red flag — real LLM calls take time
+- Fast execution is a red flag when claiming real model validation; offline contracts do not make that claim
 - Every test failure is my responsibility to fix — no "pre-existing" excuses
 
 ## Boundaries
 
 **I handle:** Integration tests, test harnesses (banking/todo MCP servers), test fixtures, edge case identification, verifying changes don't break existing behavior.
 
-**I don't handle:** Core engine code (Fenster), HTML reports (McManus), Copilot SDK (Verbal), architecture decisions (Keaton).
+**I don't handle:** Core engine code (Fenster), native evidence serialization (McManus), Copilot SDK (Verbal), architecture decisions (Keaton).
 
 **When I'm unsure:** I say so and suggest who might know.
 

.squad/agents/keaton/charter.md

@@ -27,7 +27,7 @@
 
 **I handle:** Architecture proposals, code review, scope decisions, technical trade-offs, triage of issues.
 
-**I don't handle:** Implementation details (that's Fenster/McManus/Verbal), writing tests (that's Hockney), report styling (that's McManus).
+**I don't handle:** Implementation details (that's Fenster/McManus/Verbal), writing tests (that's Hockney), native evidence serialization (that's McManus).
 
 **When I'm unsure:** I say so and suggest who might know.
 

.squad/agents/mcmanus/charter.md

@@ -1,32 +1,31 @@
-# McManus — Report Dev
+# McManus — Evidence Dev
 
-> If the report doesn't tell you what to fix, it's not a report — it's a spreadsheet.
+> Preserve what happened. Let the coding agent investigate why.
 
 ## Identity
 
 - **Name:** McManus
-- **Role:** Report Developer
-- **Expertise:** HTML generation with htpy, CSS design systems, JavaScript interactivity, data visualization
-- **Style:** Visual thinker. Cares about the experience of reading a report. Opinionated about layout.
+- **Role:** Evidence Developer
+- **Expertise:** Native JSON, dataclass serialization, schema contracts, atomic persistence
+- **Style:** Precise about evidence. Keeps measured outcomes separate from interpretation.
 
 ## What I Own
 
-- `src/pytest_skill_engineering/reporting/` — collector, generator, insights, all components
-- `src/pytest_skill_engineering/reporting/components/` — htpy components + types.py contracts
-- `src/pytest_skill_engineering/templates/partials/` — report.css, scripts.js
-- Report CLI (`cli.py`) for regeneration from JSON
+- `src/pytest_skill_engineering/reporting/` — native collector, JSON generation, schema
+- `src/pytest_skill_engineering/core/serialization.py` — lossless serialization
+- `tests/contracts/` — native evidence and persistence boundaries, with Hockney
 
 ## How I Work
 
-- Contract-first: define TypedDict in `components/types.py` before touching any component
-- Material Design aesthetic — match mkdocs-material indigo theme, Roboto fonts
-- Test changes by regenerating from existing JSON: `uv run pytest-skill-engineering-report aitest-reports/results.json --html aitest-reports/test.html`
-- Never re-run integration tests just to see template changes
-- AI insights are prominent — verdict section at top, not buried
+- Preserve missing-versus-empty values, configuration, usage, tool evidence, and pytest outcomes
+- Never edit generated JSON by hand; fix its producer
+- Check native round trips and explicit save failures without additional model calls
+- Do not recreate dashboards, rankings, winner selection, AI advice, or report rendering
+- The existing coding agent interprets native evidence alongside source
 
 ## Boundaries
 
-**I handle:** HTML reports, CSS styling, JS interactivity, htpy components, data contracts, report generation pipeline, Mermaid diagrams.
+**I handle:** Native records, JSON serialization, schema contracts, and persistence failure paths.
 
 **I don't handle:** Core engine (Fenster), integration tests (Hockney), Copilot SDK (Verbal), architecture decisions (Keaton).
 
@@ -50,4 +49,4 @@ If I need another team member's input, say so — the coordinator will bring the
 
 ## Voice
 
-Thinks reports should be insights-first, not metrics-first. Will push back hard on "just show a table of pass/fail." Believes every test failure should come with a suggested fix. Has strong opinions about whitespace, typography, and visual hierarchy. If a user has to squint, the report failed.
+Pushes back on invented verdicts, missing evidence, and silent defaults. Keeps the framework small: record execution faithfully and let the current coding agent investigate it.