Release v0.5.0

.gitattributes

@@ -0,0 +1,2 @@
+# Golden fixtures are byte-exact — never EOL-normalize them.
+tests/golden/** -text

.github/PULL_REQUEST_TEMPLATE/harness-descriptor.md

@@ -0,0 +1,65 @@
+<!--
+Harness-descriptor contribution PR — data-only.
+Use for: adding a new built-in harness descriptor (or extending one) that only
+reuses EXISTING named capabilities. New parsers, slug capabilities, shadow
+preflights, and guard support are CODE contributions — one capability per PR,
+separately from this one (see docs/progressive-enhancements.md "Guardrails").
+Guide: docs/byoh.md "Upstreaming your descriptor".
+Title: feat(harness): add <label> descriptor
+-->
+
+## Harness
+
+- **Label:** `<label>`
+- **Harness CLI + version verified against:** <!-- e.g. `cool-cli 1.4.2` -->
+- **What it is / docs link:**
+
+## Scope check (data-only)
+
+- [ ] This PR adds/changes descriptor **data only** — no parser, slug, shadow,
+      or guard code (those are separate one-capability-per-PR contributions)
+- [ ] `[transcript]`/`[shadow]`/`[staging]` (if declared) reuse an existing named
+      capability (`claude-stream-json` / `codex-items` / `opencode` / `claude-plugins`)
+      or ingest through a `[transcript.extract]` block (declarative data)
+- [ ] No `[guard]` table and no `run.supports_guard = true` in this PR
+
+## The diff (expected shape)
+
+- [ ] `harnesses/<label>.toml` — the descriptor
+- [ ] `src/adapters/descriptor.rs` — the `EMBEDDED_DESCRIPTORS` entry (mechanical
+      registration only)
+- [ ] `docs/<label>-notes.md` — verified implementation notes (promote your
+      project's `.eval-magic/harnesses/<label>-notes.md` — `harness init`
+      scaffolds it)
+- [ ] `README.md` — new support-table row (only wired enhancements checked;
+      ❌ elsewhere)
+
+## Verification evidence
+
+- [ ] Lint is clean — paste the output:
+
+  ```
+  $ eval-magic harness lint harnesses/<label>.toml
+  ```
+
+- [ ] A smoke eval ran end-to-end on this harness
+      (`run --harness <label>` → dispatch → `ingest` → `finalize`); describe or
+      link the result:
+- [ ] Every declared enhancement was exercised by that run (e.g. staged skill
+      discovered natively, events file parsed by the named parser, model flag
+      appeared in the generated recipes)
+
+## Don't-guess attestation
+
+Every flag, filename, and event shape in the descriptor is traced to the
+harness's own documentation or to output I observed — nothing is inferred by
+analogy with another harness. `docs/<label>-notes.md` records, per value: the
+value, its source (docs URL / `--help` output / observed events file), and the
+harness version.
+
+- [ ] The notes file's Verification record covers every non-comment field in
+      the descriptor
+
+## Related
+
+Issues/PRs:

AGENTS.md

@@ -4,9 +4,11 @@
 
 `eval-magic` is a Rust CLI crate. The binary entry point is `src/main.rs`; reusable logic lives in
 `src/lib.rs` and submodules such as `cli/`, `pipeline/`, `sandbox/`, `validation/`, and
-`workspace/`. JSON schemas are tracked in `schema/`, harness profiles in `profiles/`, and design or
-parity notes in `docs/`. Integration tests are split by surface area under `tests/cli/` and
-`tests/run/`; unit tests usually live beside the module they exercise.
+`workspace/`. JSON schemas are tracked in `schema/`, harness descriptors in `harnesses/`, shared
+prompt profiles in `profiles/`, and development docs (the harness enhancement contract,
+per-harness notes) in `docs/`. Integration
+tests are split by surface area under `tests/cli/` and `tests/run/`; unit tests usually live
+beside the module they exercise.
 
 ## Build, Test, and Development Commands
 

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "eval-magic"
-version = "0.4.1"
+version = "0.5.0"
 edition = "2024"
 description = "One-stop CLI for running skill evals — measure whether an agent skill actually shifts behavior."
 license = "MIT"
@@ -44,6 +44,10 @@ regex = { version = "1.12.3", default-features = false, features = ["std", "perf
 serde = { version = "1.0.228", features = ["derive"] }
 serde_json = { version = "1.0.150", features = ["preserve_order"] }
 thiserror = "2.0.18"
+# Harness descriptor files (harnesses/*.toml). `display` serializes the
+# resolved (layer-merged) descriptor back to authorable TOML for
+# `harness show`.
+toml = { version = "0.9", default-features = false, features = ["parse", "serde", "display"] }
 walkdir = "2.5.0"
 
 [dev-dependencies]