update with coverage report info

Author: NSoiffer

AGENTS.md

@@ -41,4 +41,5 @@ but add common mistakes of AI agents here instead.
 - **CI:** `.github/workflows/fuzz.yml` builds `fuzz_target_1` with default and `--features no-unsafe`, runs with `mathml.dict` and **`corpus/fuzz_target_1`** (repo: `fuzz/corpus/fuzz_target_1`). The corpus is **restored/saved via GitHub Actions cache** so inputs improve across runs. Push/PR: `-max_total_time=60` each; **weekly schedule: 3600s (1 hour) per configuration**. Weekly runs also **minimize the corpus** (`cargo fuzz cmin`) to keep the cache lean.
 - **Crash artifacts:** uploaded as GitHub Actions artifacts (90-day retention) on any fuzz failure. Download them from the failed workflow run's "Artifacts" section.
 - **Regression tests from crashes:** run `python PythonScripts/fuzz_to_test.py` (optionally pass artifact paths; defaults to `fuzz/artifacts/fuzz_target_1/`). This appends `#[test]` functions to `tests/fuzz_regressions.rs`, deduplicating by content hash. Tests call `set_mathml` + `get_spoken_text` + `get_braille` on the crash input.
+- **Coverage:** weekly CI runs `cargo fuzz coverage` against the corpus, then produces an HTML report via `llvm-cov show`. The report is uploaded as the **`fuzz-coverage-report`** Actions artifact (30-day retention). Download and open `index.html` to see which source lines the fuzzer has reached.
 - **Windows:** prefer **WSL** for local `cargo fuzz` (ASAN/libFuzzer); CI runs on `ubuntu-latest`.