v1.4.0 — real trace grading

Audit-driven trust-fix release.

Fixed

• Trace grader was vacuous: the runner only emitted token spans, so all 4 rubrics scored 100 on every run. It now translates Claude Code tool events into FILE_EDIT/read/SHELL_COMMAND spans (correlating Bash exit codes), with path relativization through symlinked workspaces. Validated against a real fast-check run.
• -j N was a silent no-op without --parallel; now enables parallel mode, and crashed parallel tasks are recorded as FAILs instead of vanishing.
• no_out_of_scope_edits honors tests/ directory entries in files_to_examine.

Added

• Baselines carry per-run trace_grade + submission readiness/trace_summary (null for span-less traces, no fake 100s). Published claude-code-custom-1.4.0-fast-check.json ships real discriminating grades.
• docs/SECURITY.md: shell-exec trust boundary + scoped per-task Docker isolation.

Changed

• Aider is a real adapter; runtime deps exact-pinned; invariant guard tests for storefront drift.

274 tests, ruff clean.

v1.3.0 — Storefront and Trust

Highlights (full entry in CHANGELOG.md)

Five-phase storefront + trust release.

Added

• `--format json` on `awb gap`, `awb compare`, `awb stability`, `awb leaderboard --readiness`
• `awb leaderboard --readiness --explain` Rich Panel with 7 sub-scores per tool
• `awb validate` one-line summary (`-v` for per-file)
• `is_stub` adapter attribute + stub fail-fast at startup
• Real `aider` adapter implementation (still `is_stub = True` pending validation)
• `RunEnvironment` provenance: `python_version`, `awb_version`, `adapter_version`, `pip_freeze_hash`
• `RunError` typed exception capture in `RunOutcome.error`
• All 100 task YAMLs backfilled with `provenance`, `contamination_risk`, `label`
• Visual contract in `awb/commands/_shared.py`
• `CITATION.cff` + `codemeta.json`
• `METHODOLOGY.md` Related Work (HAL, SWE-bench, SWE-bench Pro, LiveCodeBench, METR, Aider Polyglot, Cohen)
• GitHub Pages leaderboard workflow

Fixed

• JSONL append race under `--parallel` (`fcntl.LOCK_EX` + regression test)
• Git operations had no timeout (now `asyncio.wait_for` + per-op timeout kwarg)
• `pi` adapter blocked the event loop (now `asyncio.to_thread`)
• Doc count drift across README/METHODOLOGY/ADAPTER/ARCHITECTURE
• Misquoted Stack Overflow stat + broken SWE-bench Pro link

Changed

• Capability profile renders Unicode block bars (█░) plus `(n, conf=...)` annotation
• `awb compare` collapsed from 9 columns to 6 with delta colors
• README hero hook tightened; em-dash cadence trimmed
• Softened "first to" framing; concedes HAL and Artificial Analysis as adjacent prior art

Removed

• 29 hand-painted Pillow-synthesized demo GIFs (3.2 MB) replaced by one real `vhs`-recorded `demos/hero.gif` (297 KB)

Tests

• 246 passing (was 240 in v1.2)