GitHub Action: auto-sync parameters_habitat_rules.yaml from link

[NewGraphEnvironment]

Goal

Auto-sync inst/extdata/parameters_habitat_rules.yaml from link whenever link's default bundle changes. Eliminates the manual cp instruction in the file's header comment that's been the source of repeated drift.

Why

The bundled sample parameters_habitat_rules.yaml is fresh's fallback for frs_habitat() callers without a config. The canonical generator is lnk_rules_build() in link, which produces the YAML from link/inst/extdata/configs/default/dimensions.csv. fresh's copy is meant to track link's default bundle but has drifted twice:

• 2026-04-12 → 2026-04-28 — last sync, missing in_waterbody, area_only, polygon-edge filters, explicit edge codes. Caught + fixed in fresh 0.24.1.
• The next drift will happen the next time link's default bundle evolves.

Manual cp doesn't survive contact with day-to-day work. Wanted: a workflow that does the sync without anyone remembering.

Pattern to mirror

link already has .github/workflows/sync-bcfishpass-csvs.yml — a daily cron that pulls upstream bcfishpass CSVs into link with auto-PR + auto-merge for byte-only drift, halt on shape drift. Same shape works here, just inverted: fresh pulls FROM link instead of bcfishpass.

Proposed workflow

.github/workflows/sync-rules-yaml.yml in fresh:

• Trigger: workflow_dispatch (manual) + on push to link's main affecting inst/extdata/parameters_habitat_rules.yaml (via repository_dispatch from a link-side workflow). Optionally also a daily cron as a safety net.
• Steps:
  1. Check out fresh.
  2. Fetch link's current inst/extdata/parameters_habitat_rules.yaml (raw URL on main).
  3. Compare to fresh/inst/extdata/parameters_habitat_rules.yaml.
  4. If different: open a PR with the new YAML + a NEWS-friendly diff summary (which species changed, which predicates added/removed). Auto-merge if no shape change (same predicate names); halt for shape drift, label schema-drift for coordinated review.
  5. If same: no-op.

Where crate fits

crate is being scoped as the data-ingestion layer between external sources (bcfishpass, fwapg, FOI) and downstream packages. The rules.yaml sync is sibling-package coordination, NOT external data ingestion — different concern. crate doesn't need to handle this.

If/when crate matures into a more general "config provenance" tool (e.g., it tracks shape fingerprints + signs off on inter-package config moves), this sync could route through it. But for now, a focused GitHub Action is the right shape — match the existing sync-bcfishpass-csvs.yml pattern, no new package.

Scope

Phase 1 — basic auto-sync

• workflow_dispatch + daily cron trigger
• raw URL fetch + diff + auto-PR
• auto-merge on byte-equal-after-normalization (just whitespace differences) or actual no-op
• label + halt on shape drift

Phase 2 — link-side push trigger

• link adds a workflow that on push to main affecting parameters_habitat_rules.yaml, fires a repository_dispatch to fresh
• fresh's sync workflow listens to repository_dispatch + runs immediately
• closes the latency gap (cron lag → seconds)

Phase 3 — defer

• Multi-bundle support: sync per-bundle rules.yaml files into named fresh fixtures (fresh/inst/testdata/rules-bcfishpass.yaml, rules-default.yaml). Useful for fresh's tests but not required.
• Provenance tracking: write the link commit SHA into the synced file's header comment so drift is debuggable.

Out of scope

• Renaming or relocating fresh's bundled YAML.
• Removing the manual cp instruction from the YAML header until phase 1 lands and proves itself.

Test plan

• Manual workflow_dispatch: produces a PR if drift exists; no-op if not.
• Synthetic drift test: inject a small change in link's parameters_habitat_rules.yaml on a fork branch, point the workflow at the fork's raw URL, confirm PR opens with the right body.
• Shape-drift test: inject a change that adds a new predicate to the YAML structure; confirm the workflow labels the PR schema-drift and doesn't auto-merge.