Free, open source blog writer for Claude Code. Researches current sources, writes answer-first posts for Google and AI answers, has an independent reviewer score each one out of 100 and ships only after five delivery gates. 32 skills, 30 /blog commands, 5 agents. Active development at AI-Marketing-Hub/claude-blog; public releases ship here.
Python agent-skillsaiai-citationsai-contentai-marketingai-marketing-hubblogblog-writingclaude-codeclaude-code-skillclaude-pluginclaude-skillcontent-creationcontent-optimizationcontent-strategyeeatgeomultilingualopen-sourceseo
"""Real YAML metadata regressions and installed-relative Markdown contracts."""
from __future__ import annotations
import os
from pathlib import Path
import re
import subprocess
import sys
import pytest
ROOT = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(ROOT / "scripts"))
from validate_skills import FrontmatterError, parse_frontmatter, validate_bundle, validate_frontmatter
def skill(**updates):
fields = dict(name="blog-fixture", description="Fixture", license="MIT", **{"user-invocable": True, "argument-hint": "<file>"})
fields.update(updates)
return fields
def test_maintained_bundle_uses_one_yaml_and_policy_validator():
assert validate_bundle(ROOT) == []
values = {path.parent.name: parse_frontmatter(path.read_text())["user-invocable"] for path in (ROOT / "skills").glob("*/SKILL.md")}
assert len(values) == 32
assert values.pop("blog-chart") is False
assert list(values.values()) == [True] * 31
assert len(list((ROOT / "agents").glob("*.md"))) == 5
@pytest.mark.parametrize("yaml_body", ["name: one\nname: two", "metadata:\n version: a\n version: b", "description: [unterminated", "- one\n- two", "null", "1: value"])
def test_real_yaml_rejects_duplicate_malformed_and_nonmapping_frontmatter(yaml_body):
with pytest.raises(FrontmatterError):
parse_frontmatter(f"---\n{yaml_body}\n---\nBody")
@pytest.mark.parametrize("text", ["Body", "\n---\nname: one\n---", "---\nname: one"])
def test_missing_or_unterminated_yaml_fails(text):
with pytest.raises(FrontmatterError):
parse_frontmatter(text)
def test_folded_description_map_metadata_and_yaml_booleans():
fields = parse_frontmatter('---\nname: blog-fixture\ndescription: >\n First line\n second line\nlicense: MIT\nuser-invocable: false\nmetadata:\n version: "1.0"\n nested: {reviewed: true}\n---\nBody')
assert "First line second line" in fields["description"]
assert fields["metadata"]["nested"]["reviewed"] is True
assert validate_frontmatter(fields) == []
def test_native_missing_invocation_defaults_true_but_project_requires_explicit():
fields = skill()
del fields["user-invocable"]
assert validate_frontmatter(fields, project_policy=False) == []
assert any("host defaults to true" in error for error in validate_frontmatter(fields))
fields["user-invokable"] = False
assert any("misspelled" in error for error in validate_frontmatter(fields))
@pytest.mark.parametrize("updates", [{"user-invocable": "false"}, {"user-invocable": 0}, {"metadata": "version: 1"}, {"description": ["words"]}, {"allowed-tools": {"Bash": True}}, {"disable-model-invocation": []}])
def test_project_typed_metadata_rejects_wrong_shapes(updates):
assert any("invalid type" in error for error in validate_frontmatter(skill(**updates)))
@pytest.mark.parametrize("tools", ["Read Grep", "Read, Grep", ["Read", "Grep", "Bash(git status *)"]])
def test_allowed_tools_host_support_is_separate_from_no_preapproval_policy(tools):
fields = skill(**{"allowed-tools": tools})
assert validate_frontmatter(fields, project_policy=False) == []
assert validate_frontmatter(fields) == ["allowed-tools is valid host preapproval metadata, but project policy forbids skill preapprovals"]
@pytest.mark.parametrize("tools", ["Read, Grep, Bash", ["Read", "Bash"], ["Bash(git status *)"]])
def test_agent_actual_allowlist_rejects_bash_grants(tools):
fields = dict(name="fixture", description="Fixture", tools=tools)
assert any("Bash" in error for error in validate_frontmatter(fields, agent=True))
assert validate_frontmatter(fields, agent=True, project_policy=False) == []
def test_agent_explicit_tools_policy_does_not_silently_inherit():
assert validate_frontmatter(dict(name="fixture", description="Fixture"), agent=True) == ["missing required field tools"]
def test_cli_nonzero_diagnostics_for_real_invalid_file(tmp_path):
path = tmp_path / "skills" / "blog-fixture" / "SKILL.md"
path.parent.mkdir(parents=True)
path.write_text("---\nname: first\nname: duplicate\n---\n")
result = subprocess.run([sys.executable, str(ROOT / "scripts/validate_skills.py"), "--root", str(tmp_path)], capture_output=True, text=True)
assert result.returncode == 1
assert "skills/blog-fixture/SKILL.md: duplicate field 'name'" in result.stderr
@pytest.fixture(params=["plugin", "standalone"])
def installed_layout(tmp_path, request):
package = tmp_path / ("cached plugin version" if request.param == "plugin" else "standalone profile")
anchor = package / "skills" / "blog-style"
anchor.mkdir(parents=True)
helpers = package / "scripts"
helpers.mkdir()
(helpers / "style_learn.py").write_text("from pathlib import Path\nprint(Path(__file__).resolve())\n")
reference = package / "skills/blog/references/quality-scoring.md"
reference.parent.mkdir(parents=True)
reference.write_text("TRUSTED BUNDLED REFERENCE")
cwd = tmp_path / "unrelated consumer"
(cwd / "scripts").mkdir(parents=True)
(cwd / "scripts/style_learn.py").write_text("raise AssertionError('POISON CWD')")
(cwd / "skills/blog/references").mkdir(parents=True)
(cwd / "skills/blog/references/quality-scoring.md").write_text("POISON CWD")
return anchor, helpers, reference, cwd
def run_style_snippet(anchor, cwd, **overrides):
# The host renders this placeholder in Markdown. No shell env variable is
# supplied, so accidental dependence on an exported host variable fails.
text = (ROOT / "skills/blog-style/SKILL.md").read_text()
block = re.search(r"```bash\n(.*?)\n```", text, re.S).group(1)
block = block.replace("${CLAUDE_SKILL_DIR}", str(anchor)).replace("<paths>", "fixture.md")
return run_rendered_block(block, cwd, **overrides)
def run_rendered_block(block, cwd, **overrides):
env = {key: value for key, value in os.environ.items() if not key.startswith("CLAUDE_BLOG_") and key != "CLAUDE_SKILL_DIR"}
env.update(overrides)
env["PATH"] = str(Path(sys.executable).parent) + os.pathsep + env.get("PATH", "")
return subprocess.run(["bash", "-c", block], cwd=cwd, env=env, capture_output=True, text=True)
def test_rendered_installed_paths_choose_package_helper_and_reference(installed_layout):
anchor, helpers, reference, cwd = installed_layout
result = run_style_snippet(anchor, cwd)
assert result.returncode == 0, result.stderr
assert result.stdout.strip() == str(helpers / "style_learn.py")
text = (ROOT / "skills/blog-write/SKILL.md").read_text()
ref = re.search(r"`(\$\{CLAUDE_SKILL_DIR\}/../blog/references/quality-scoring.md)`", text).group(1)
assert Path(ref.replace("${CLAUDE_SKILL_DIR}", str(anchor))).resolve() == reference
assert Path(ref.replace("${CLAUDE_SKILL_DIR}", str(anchor))).read_text() == "TRUSTED BUNDLED REFERENCE"
def test_missing_installed_helper_fails_without_cwd_fallback(installed_layout):
anchor, helpers, _, cwd = installed_layout
(helpers / "style_learn.py").unlink()
result = run_style_snippet(anchor, cwd)
assert result.returncode != 0
assert "POISON CWD" not in result.stderr
def test_operator_absolute_override_precedence_and_quoted_metacharacters(installed_layout, tmp_path):
anchor, _, _, cwd = installed_layout
override = tmp_path / "trusted override $(touch INJECTED)"
override.mkdir()
(override / "style_learn.py").write_text("print('OPERATOR OVERRIDE')")
result = run_style_snippet(anchor, cwd, CLAUDE_BLOG_SCRIPTS_DIR=str(override))
assert result.returncode == 0, result.stderr
assert result.stdout.strip() == "OPERATOR OVERRIDE"
assert not (cwd / "INJECTED").exists()
@pytest.mark.parametrize("override", ["scripts", "../scripts", "$(touch INJECTED)", "style_learn.py; touch INJECTED"])
def test_relative_and_malicious_helper_overrides_fail(installed_layout, override):
anchor, _, _, cwd = installed_layout
result = run_style_snippet(anchor, cwd, CLAUDE_BLOG_SCRIPTS_DIR=override)
assert result.returncode == 1
assert "script dir must be absolute" in result.stderr
assert not (cwd / "INJECTED").exists()
def test_context_loader_uses_resolved_core_root_and_helper_override(installed_layout, tmp_path):
anchor, helpers, _, cwd = installed_layout
text = (ROOT / "skills/blog/SKILL.md").read_text()
block = next(block for block in re.findall(r"```bash\n(.*?)\n\s*```", text, re.S) if 'if [ -n "${CLAUDE_BLOG_LOAD_UNTRUSTED_HELPER:-}" ]' in block)
block = block.replace("${CLAUDE_SKILL_DIR}", str(anchor.parent / "blog"))
loader = helpers / "load_untrusted_root.py"
loader.write_text("from pathlib import Path\nprint(Path(__file__).resolve())")
result = run_rendered_block(block, cwd)
assert result.returncode == 0, result.stderr
assert result.stdout.strip() == str(loader)
override = tmp_path / "trusted loader.py"
override.write_text("print('EXPLICIT TRUSTED LOADER')")
result = run_rendered_block(block, cwd, CLAUDE_BLOG_LOAD_UNTRUSTED_HELPER=str(override))
assert result.returncode == 0, result.stderr
assert result.stdout.strip() == "EXPLICIT TRUSTED LOADER"
result = run_rendered_block(block, cwd, CLAUDE_BLOG_LOAD_UNTRUSTED_HELPER="scripts/load_untrusted_root.py")
assert result.returncode == 1
assert "helper path must be absolute" in result.stderr
loader.unlink()
(cwd / "scripts/load_untrusted_root.py").write_text("raise AssertionError('POISON CWD')")
result = run_rendered_block(block, cwd)
assert result.returncode == 1
assert "trusted load_untrusted_root.py not found" in result.stderr
@pytest.mark.parametrize("plugin", [True, False])
def test_update_ledger_selection_uses_installed_ancestry_not_cwd(tmp_path, plugin):
package = tmp_path / "trusted package with spaces"
anchor = package / "skills/blog"
anchor.mkdir(parents=True)
if plugin:
manifest = package / ".claude-plugin/plugin.json"
manifest.parent.mkdir()
manifest.write_text('{"name":"fixture"}')
ledger = package / "data/google-updates.json"
else:
ledger = anchor / "data/google-updates.json"
ledger.parent.mkdir()
ledger.write_text("TRUSTED LEDGER")
cwd = tmp_path / "unrelated consumer"
(cwd / "data").mkdir(parents=True)
(cwd / "data/google-updates.json").write_text("POISON LEDGER")
text = (ROOT / "skills/blog/SKILL.md").read_text()
block = next(block for block in re.findall(r"```bash\n(.*?)\n```", text, re.S) if 'BLOG_UPDATE_LEDGER=' in block)
block = block.replace("${CLAUDE_SKILL_DIR}", str(anchor)) + '\ncat "$BLOG_UPDATE_LEDGER"'
result = run_rendered_block(block, cwd)
assert result.returncode == 0, result.stderr
assert result.stdout == "TRUSTED LEDGER"
ledger.unlink()
result = run_rendered_block(block, cwd)
assert result.returncode == 1
assert "installed update ledger missing" in result.stderr
def test_own_script_paths_and_sibling_overrides_are_host_rendered(installed_layout, tmp_path):
anchor, _, _, cwd = installed_layout
google = anchor.parent / "blog-google"
(google / "scripts").mkdir(parents=True)
runner = google / "scripts/run.py"
runner.write_text("from pathlib import Path\nprint(Path(__file__).resolve())")
text = (ROOT / "skills/blog-google/SKILL.md").read_text()
block = next(block for block in re.findall(r"```bash\n(.*?)\n```", text, re.S) if "BLOG_SKILLS_DIR=" in block)
block = block.replace("${CLAUDE_SKILL_DIR}", str(google))
result = run_rendered_block(block, cwd)
assert result.returncode == 0, result.stderr
assert result.stdout.strip() == str(runner)
result = run_rendered_block(block, cwd, CLAUDE_BLOG_SKILLS_DIR="skills")
assert result.returncode == 1
assert "skills dir must be absolute" in result.stderr
override = tmp_path / "trusted skills override"
(override / "blog-google/scripts").mkdir(parents=True)
(override / "blog-google/scripts/run.py").write_text("print('TRUSTED SIBLING OVERRIDE')")
result = run_rendered_block(block, cwd, CLAUDE_BLOG_SKILLS_DIR=str(override))
assert result.returncode == 0, result.stderr
assert result.stdout.strip() == "TRUSTED SIBLING OVERRIDE"