stage: metabolomics/v1 (initial dogfooding cut) #19

Workflow file for this run

.github/workflows/validate.yml at d7c3aaf

	# validate.yml -- runs on every PR
	# Enforces CI gates per section 14 of ASB-Skills Release Design Doc v2.
	#
	# Gates implemented here:
	# 1. LinkML schema validation (collection.yaml, tools/*.yaml)
	# 2. No orphan skills (DOI resolution sample)
	# 5. Description discipline lint (leading phrase, length, no marketing)
	# 6. EDAM IRI resolution
	# 8. RO-Crate validation (Workflow Run Profile 0.5)
	# 9. Indicium round-trip (verify-claims CLI from indicium-adapters)
	# 10. Plugin manifest validation

	name: Validate

	on:
	pull_request:
	branches: [main]
	push:
	branches: [main]

	permissions:
	contents: read
	pull-requests: write

	jobs:
	validate:
	runs-on: ubuntu-latest
	steps:
	- name: Checkout
	uses: actions/checkout@v4

	- name: Set up Python 3.12
	uses: actions/setup-python@v5
	with:
	python-version: "3.12"

	- name: Install dependencies
	run: \|
	pip install --upgrade pip
	pip install pyyaml jsonschema requests
	pip install linkml linkml-runtime \|\| echo "WARNING: linkml install failed"
	pip install rocrate \|\| echo "WARNING: rocrate install failed"
	# indicium-adapters provides the verify-claims console script (Plan 1c)
	pip install indicium-adapters \|\| echo "WARNING: indicium-adapters not yet available; gate 9 will warn-only"

	# -- Gate 10: Plugin manifest validates --------------------------------
	- name: Validate .claude-plugin/marketplace.json
	run: \|
	python - <<'EOF'
	import json, pathlib, sys
	p = pathlib.Path(".claude-plugin/marketplace.json")
	if not p.exists():
	print("SKIP: no marketplace.json found"); sys.exit(0)
	data = json.loads(p.read_text())
	required = ["schema_version", "plugins"]
	missing = [k for k in required if k not in data]
	if missing:
	print(f"FAIL: marketplace.json missing keys: {missing}"); sys.exit(1)
	if not isinstance(data["plugins"], list):
	print("FAIL: plugins must be a list"); sys.exit(1)
	print(f"PASS: marketplace.json valid ({len(data['plugins'])} plugins)")
	EOF

	# -- Gate 5: Description discipline lint --------------------------------
	- name: Lint skill descriptions
	run: \|
	python - <<'EOF'
	import sys, pathlib, yaml

	APPROVED_PREFIXES = (
	"Use when", "Reference for", "Explains", "Decision support for"
	)
	MIN_LEN = 50
	MAX_LEN = 300
	MARKETING_TERMS = ["best", "state-of-the-art", "revolutionary", "leading", "superior"]

	failures = []
	skill_files = list(pathlib.Path("collections").rglob("SKILL.md"))
	skill_files += list(pathlib.Path("staged-collections").rglob("SKILL.md"))
	for skill_md in skill_files:
	text = skill_md.read_text()
	if not text.startswith("---"):
	continue
	try:
	parts = text.split("---", 2)
	fm = yaml.safe_load(parts[1])
	except Exception:
	continue
	desc = (fm.get("description") or "").strip()
	if not desc:
	failures.append(f"{skill_md}: missing description")
	continue
	if not any(desc.startswith(p) for p in APPROVED_PREFIXES):
	failures.append(
	f"{skill_md}: description must start with one of {APPROVED_PREFIXES}"
	)
	if len(desc) < MIN_LEN:
	failures.append(
	f"{skill_md}: description too short ({len(desc)} < {MIN_LEN})"
	)
	if len(desc) > MAX_LEN:
	failures.append(
	f"{skill_md}: description too long ({len(desc)} > {MAX_LEN})"
	)
	for term in MARKETING_TERMS:
	if term.lower() in desc.lower():
	failures.append(
	f"{skill_md}: marketing term '{term}' in description"
	)
	if failures:
	print("FAIL: description discipline violations:")
	for f in failures:
	print(f" - {f}")
	sys.exit(1)
	print(f"PASS: description discipline OK ({len(skill_files)} skill files checked)")
	EOF

	# -- Gate 2: No orphan skills (DOI resolution sample) ------------------
	- name: Check derived_from DOIs resolve (sample)
	run: \|
	python - <<'EOF'
	import sys, pathlib, yaml, urllib.request, urllib.error

	failures = []
	checked = 0
	skill_files = list(pathlib.Path("collections").rglob("SKILL.md"))
	skill_files += list(pathlib.Path("staged-collections").rglob("SKILL.md"))
	# Skip auto-generated router skills — they aggregate, not derive
	# from a single paper.
	skill_files = [s for s in skill_files if s.parent.name != "_router"]
	for skill_md in skill_files[:10]:
	text = skill_md.read_text()
	if not text.startswith("---"):
	continue
	try:
	fm = yaml.safe_load(text.split("---", 2)[1])
	except Exception:
	continue
	# Accept both canonical derived_from and the upstream
	# provenance.source_papers shape used by the ASB pipeline.
	derived = fm.get("derived_from") or []
	if not derived:
	prov = fm.get("provenance") or {}
	if isinstance(prov, dict):
	src = prov.get("source_papers") or []
	derived = [s for s in src if isinstance(s, dict) and s.get("doi")]
	if not derived:
	failures.append(f"{skill_md}: no derived_from DOIs")
	continue
	# Sample: check first DOI only to keep CI fast.
	# Use doi.org's content-negotiation API (Accept: citeproc+json)
	# to validate the DOI exists, bypassing publisher-side
	# firewalls that 403 generic HEAD requests.
	entry = derived[0]
	doi = entry.get("doi") if isinstance(entry, dict) else entry
	url = f"https://doi.org/{doi}"
	try:
	req = urllib.request.Request(
	url,
	headers={
	"Accept": "application/citeproc+json",
	"User-Agent": "asb-skill-collections/0.1 (mailto:louisfelix.nothias@gmail.com)",
	},
	)
	with urllib.request.urlopen(req, timeout=15) as resp:
	# 200 with JSON body confirms the DOI exists in the
	# Crossref/DataCite registry.
	if resp.status == 200:
	checked += 1
	else:
	failures.append(
	f"{skill_md}: DOI {doi} returned status {resp.status}"
	)
	except urllib.error.HTTPError as e:
	if e.code == 404:
	failures.append(f"{skill_md}: DOI {doi} not found (404)")
	else:
	# Treat 4xx (excluding 404) and 5xx as soft-fail —
	# doi.org's content-neg API is occasionally rate-limited
	# or down; non-404 errors shouldn't block the release.
	print(f"WARN: {skill_md}: DOI {doi} soft-fail HTTP {e.code} (not blocking)")
	checked += 1
	except Exception as e:
	print(f"WARN: {skill_md}: DOI {doi} soft-fail {type(e).__name__} (not blocking)")
	checked += 1
	if failures:
	print("FAIL: orphan skill / DOI resolution failures:")
	for f in failures:
	print(f" - {f}")
	sys.exit(1)
	print(f"PASS: DOI resolution OK ({checked} DOIs checked)")
	EOF

	# -- Gate 6: EDAM IRI resolution ----------------------------------------
	- name: Check EDAM IRIs
	run: \|
	python - <<'EOF'
	import sys, pathlib, yaml

	EDAM_BASE = "http://edamontology.org/"
	failures = []
	checked = set()
	skill_files = list(pathlib.Path("collections").rglob("SKILL.md"))
	skill_files += list(pathlib.Path("staged-collections").rglob("SKILL.md"))
	for skill_md in skill_files:
	text = skill_md.read_text()
	if not text.startswith("---"):
	continue
	try:
	fm = yaml.safe_load(text.split("---", 2)[1])
	except Exception:
	continue
	meta = fm.get("metadata") or {}
	iris = []
	if meta.get("edam_operation"):
	iris.append(meta["edam_operation"])
	iris.extend(meta.get("edam_topics") or [])
	for iri in iris:
	if iri in checked:
	continue
	checked.add(iri)
	if not iri.startswith(EDAM_BASE):
	failures.append(
	f"{skill_md}: EDAM IRI {iri} does not start with {EDAM_BASE}"
	)
	if failures:
	print("FAIL: EDAM IRI violations:")
	for f in failures:
	print(f" - {f}")
	sys.exit(1)
	print(f"PASS: EDAM IRIs OK ({len(checked)} unique IRIs validated)")
	EOF

	# -- Gate 8: RO-Crate validation ----------------------------------------
	- name: Validate RO-Crate metadata
	run: \|
	python - <<'EOF'
	import sys, json, pathlib

	crate_files = list(pathlib.Path("collections").rglob("ro-crate-metadata.json"))
	crate_files += list(pathlib.Path("staged-collections").rglob("ro-crate-metadata.json"))
	failures = []
	for crate_file in crate_files:
	try:
	data = json.loads(crate_file.read_text())
	if "@context" not in data:
	failures.append(f"{crate_file}: missing @context")
	if "@graph" not in data:
	failures.append(f"{crate_file}: missing @graph")
	graph = data.get("@graph", [])
	root_ids = {"./", "."}
	root_entities = [e for e in graph if e.get("@id") in root_ids]
	if not root_entities:
	failures.append(
	f"{crate_file}: no root dataset entity (id ./ or .)"
	)
	except json.JSONDecodeError as e:
	failures.append(f"{crate_file}: JSON parse error: {e}")
	if failures:
	print("FAIL: RO-Crate validation failures:")
	for f in failures:
	print(f" - {f}")
	sys.exit(1)
	print(f"PASS: RO-Crate validation OK ({len(crate_files)} crates checked)")
	EOF

	# -- Gate 9: indicium round-trip (verify-claims) -------------------------
	- name: verify-claims round-trip
	run: \|
	if ! command -v verify-claims &> /dev/null; then
	echo "WARNING: verify-claims CLI not found."
	echo "Install indicium-adapters to enable gate 9: pip install indicium-adapters"
	echo "Skipping gate 9 (non-blocking until indicium-adapters is published)"
	exit 0
	fi
	EXIT=0
	for collection_dir in collections//v; do
	[ -d "$collection_dir" ] \|\| continue
	echo "Running verify-claims on $collection_dir ..."
	verify-claims --collection "$collection_dir" --format json \|\| EXIT=$?
	done
	for collection_dir in staged-collections//v; do
	[ -d "$collection_dir" ] \|\| continue
	echo "Running verify-claims on $collection_dir ..."
	verify-claims --collection "$collection_dir" --format json \|\| EXIT=$?
	done
	exit $EXIT

	# -- Gate 1: LinkML schema validation ------------------------------------
	- name: LinkML schema validation
	run: \|
	python - <<'EOF'
	import sys, pathlib, subprocess, importlib.util

	# asb-schema is not yet on PyPI; the LinkML schema lives in the
	# AgenticScienceBuilder repo at scitask-schema/. CI here doesn't
	# have it checked out, and a wheel install isn't viable yet.
	#
	# Until asb-schema is published (v1.1+), this step performs a
	# structural sanity check on collection.yaml instead of full
	# LinkML conformance: required top-level keys + value types.
	# Real LinkML validation runs at release time once asb-schema
	# ships to PyPI.

	import yaml
	required_keys = {"slug", "title", "version", "n_skills", "skills"}
	collection_files = list(pathlib.Path("collections").rglob("collection.yaml"))
	collection_files += list(
	pathlib.Path("staged-collections").rglob("collection.yaml")
	)
	if not collection_files:
	print("SKIP: no collection.yaml files found")
	sys.exit(0)

	failures = []
	for cf in collection_files:
	try:
	doc = yaml.safe_load(cf.read_text()) or {}
	except Exception as exc:
	failures.append(f"{cf}: YAML parse error: {exc}")
	continue
	missing = required_keys - set(doc.keys())
	if missing:
	failures.append(f"{cf}: missing keys {sorted(missing)}")
	if not isinstance(doc.get("skills"), list):
	failures.append(f"{cf}: 'skills' must be a list")
	if not isinstance(doc.get("n_skills"), int):
	failures.append(f"{cf}: 'n_skills' must be int")
	if failures:
	print("FAIL: structural validation failures:")
	for f in failures:
	print(f" - {f}")
	sys.exit(1)
	print(f"PASS: structural validation OK ({len(collection_files)} files)")
	print("INFO: Full LinkML conformance deferred until asb-schema ships to PyPI (v1.1).")
	EOF

Navigation Menu

Uh oh!

stage: metabolomics/v1 (initial dogfooding cut) #19

Workflow file

stage: metabolomics/v1 (initial dogfooding cut) #19

Uh oh!

Workflow file for this run