First walk: cwl-to-galaxy end-to-end via ga4gh_challenge (earn the CWL-source tier out of draft)

_Filed by Claude (AI assistant) on behalf of @jmchilton._

The complementary direction to the → CWL-target scoping (`nextflow-to-cwl` scenarios). **CWL-source → Galaxy** (`cwl-to-galaxy`) is the *most tractable* first walk of any CWL-involving pipeline, because its downstream authoring tier is already proven.

## Why this one is tractable now

`cwl-to-galaxy` splits cleanly:

| Tier | Molds | State |
|---|---|---|
| Downstream Galaxy authoring | `compare-against-iwc-exemplar`, `advance-galaxy-draft-step`, `implement-galaxy-workflow-test`, `validate-galaxy-workflow` | **reviewed** — proven in #369 |
| CWL-source upstream + test | `summarize-cwl`, `cwl-summary-to-galaxy-{interface,data-flow,template}`, `cwl-to-test-data`, `cwl-test-to-galaxy-test-plan` | **draft, but real-bodied** — have bodies + references, just unexercised |

Contrast with the → CWL-**target** direction (`nextflow-to-cwl`/`paper-to-cwl`): those are blocked on seeded-stub molds with no bodies/oracles. Here, nothing is a stub — the whole spine exists; it just hasn't been *walked* to earn the source tier out of draft the way #369 did for the Galaxy pipelines.

## Fixture (pre-pinned for exactly this)

`Barski-Lab/ga4gh_challenge` — `fixtures.yaml` entry, tier `tiny`, entrypoint `biowardrobe_chipseq_se.cwl` (single SE ChIP-seq workflow + one subworkflow). Its own note: *"Small, self-contained, fully resolvable — good baseline for end-to-end CWL→Galaxy conversion smoke tests."* Materialize with `make fixtures-cwl`.

Alternatives considered:
- `common-workflow-language/user_guide` `1st-workflow.cwl` — tiniest, but tar-extract + `javac` has no Galaxy tool mapping, so an end-to-end conversion stalls at tool resolution. Good only as a `summarize-cwl` smoke test.
- `EBI-Metagenomics/pipeline-v5` — real bio but *large* (conditionals + subworkflows); too much for a first walk.

## Proposed deliverable

1. Author `content/pipelines/cwl-to-galaxy/{scenarios.md,eval.md}`, grounded in `ga4gh_challenge`, mirroring `nextflow-to-galaxy` (and the `nextflow-to-cwl` scaffolding just added for the other direction).
2. Attempt the basic conversion walk of `biowardrobe_chipseq_se.cwl` → gxformat2 through the real spine.
3. On the evidence, **earn the CWL-source molds out of draft** (`summarize-cwl`, `cwl-summary-to-galaxy-*`, `cwl-to-test-data`, `cwl-test-to-galaxy-test-plan`) — the #369 pattern applied to the CWL-source tier.

## Why it matters

Proves the CWL-source path, honestly promotes ~6 draft molds on produced artifacts (not metadata sweeps), and de-risks the CWL work before investing in the harder → CWL-target authoring tier. The ChIP-seq domain also has strong IWC exemplar coverage, so `compare-against-iwc-exemplar` should have real ground to stand on.

Start with the existing nextflow-to-galaxy content and the fixtures.yaml entry for Barski-Lab/ga4gh_challenge. Run make fixtures-cwl, then attempt the biowardrobe_chipseq_se.cwl to gxformat2 walk through the real spine. Done means adding content/pipelines/cwl-to-galaxy/{scenarios.md,eval.md}, producing evidence from the walk, and deciding whether the listed CWL-source molds can leave draft.

Written by the indexing model from the issue text.