rigforge v1.1.0

Polish wave (2026-10-19, same day):

• This commit removes unused imports and unused local variables in
  rigforge/telemetry.py, rigforge/tests/test_cli.py,
  rigforge/tests/test_public_attest.py, rigforge/benchmark.py,
  rigforge/cli.py, rigforge/commands/status.py, rigforge/evalloop.py,
  rigforge/lattice/router.py, rigforge/public_attest.py,
  rigforge/registry.py, rigforge/triple_diamond.py and four test modules,
  so ruff check --isolated (the full default rule set) passes clean.
• This commit changes .github/workflows/ci.yml and .github/workflows/smoke.yml
  to add timeout-minutes: 15 to every job and to remove
  continue-on-error: true from the full-rule-set lint step, which is now
  blocking because it passes.
• This commit changes the README CI badge to the shields
  github/actions/workflow/status form for ci.yml on main.
• This commit adds rigforge/tests/test_repo_hygiene.py, which asserts from the
  files on disk: .github/CODEOWNERS contains * @mrodgersjs-web;
  .github/copilot-instructions.md is at most 8000 bytes and names a test
  command; the README carries a github/actions/workflow/status badge, the
  rodgersintelligence.com footer, a test command word, and clone examples
  targeting mrodgersjs-web/rigforge; and every workflow job has
  timeout-minutes and no continue-on-error.

Lint repair wave (2026-10-20):

• This commit sorts the import blocks in the contracts/ modules and tests that ruff check --isolated flagged as unsorted (rule I001); no behavior changes.
• This commit sorts the import blocks in the top-level rigforge/ modules flagged unsorted by ruff check --isolated (rule I001); no behavior changes.
• This commit sorts the import blocks in the rigforge/tests/ modules flagged unsorted by ruff check --isolated (rule I001); no behavior changes.
• This commit sorts the import blocks in the rigforge/lattice/ and rigforge/commands/ modules flagged unsorted by ruff check --isolated (rule I001); no behavior changes.
• This commit removes the four # noqa markers that ruff check --isolated reported as unused (rule RUF100), keeping a marker's reason text as a plain comment where it had one.
• This commit applies ruff's remaining safe autofixes under the default rule set (UP035, UP037, UP045, FURB167, RUF022, PLR0402, TC005, SIM117, PLR1711, RET501); annotation and style modernization only, no behavior changes.
• This commit narrows one blind except to ValueError and keeps the eight deliberate catch-all handlers with # noqa reasons (rules BLE001, S110); no behavior changes.
• This commit asserts the specific pydantic.ValidationError in three contract tests and makes dict iteration and subprocess check=False explicit (rules B017, PLC0206, PLW1510); no behavior changes.
• This commit clears the last seven default-set findings: a merged nested if (SIM102), a tz-aware test fixture (DTZ001), a keys-only loop (PERF102), an unused unpack renamed (RUF059), a dict literal (C408), rigforge/cli.py made executable for its shebang (EXE001), and one deliberate ValueError kept with a # noqa reason (TRY004).
• This commit makes the workflow and config comments truthful: the full-default lint step's "passes it clean" claim was false (142 findings) until this repair, the test-count comment read 320 while the suite runs 325, and pyproject.toml called the full set advisory while CI runs it blocking.

Added

• Tests for the intent-routing subsystem (lattice model, intent parser, lattice
  router, troika triad, forge engine, standalone commands) and the telemetry
  no-op fallback: 25 tests, rigforge/tests/test_intent_pipeline.py.
• Packaging invariant tests (rigforge/tests/test_packaging.py) asserting the
  installed version, the py.typed marker, the console script, and the Python
  floor.
• CI config invariant tests (rigforge/tests/test_ci_config.py) that parse both
  workflow files and assert triggers, read-only permissions, the version matrix,
  and that the real suite runs.
• rigforge/py.typed (partial typing marker).
• Community files: CODE_OF_CONDUCT.md, this changelog, issue templates, a pull
  request template, and CITATION.cff.
• docs/USAGE.md: the CLI reference, configuration, MCP integration, spec-bound
  proofs, merge gate, swarm verdict board, and the public-attestation compose
  path, moved off the README front page.

Changed

• README rewritten to the front-page pattern: one-line pitch, badges, kept demo
  GIF, "Why this exists", a "What it proves" table of runnable checks with
  measured results, quickstart, real example output, a mermaid architecture
  diagram, project layout, development, and the public boundary.
• Packaging: version 1.0.0 -> 1.1.0, authors "Mike Rodgers", plain-English
  description, requires-python >=3.10 with the 3.10 classifier and ruff
  target-version = "py310", and cryptography added to the dev extra so the
  public-attestation tests run instead of skipping.
• The python_version hard gate now defaults to 3.10, matching the advertised
  floor in pyproject.toml and the CI matrix (it enforced 3.11).
• CI: the test matrix covers Python 3.10 to 3.13 and runs pytest with coverage;
  the smoke workflow runs the full test suite before scripts/smoke.sh.

Fixed

• README claimed "252 passing"; the measured result on this version is 320
  passed, 0 skipped, with 84% package coverage (test files excluded).
• Stale test-count comments in pyproject.toml and .github/workflows/ci.yml.

Built by Mike Rodgers · rodgersintelligence.com