AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary
Python ai-promptsai-skillblueskyclaudeclaude-codeclawhubdeep-researchhackernewsinstagramopenclawpolymarketrecencyredditresearchsocial-mediatiktoktrendstwitterweb-searchyoutube
"""The loaded root routes to bounded instructions that survive distribution."""
import io
import os
from pathlib import Path
import re
import shutil
import subprocess
import sys
import zipfile
import pytest
import yaml
from tests.skill_contract import (
ROUTES, SKILL_ROOT, STANDARD_CONDITIONS, reference_text, root_text, routing_table,
)
ROOT = SKILL_ROOT.parents[1]
def test_every_reference_has_the_correct_conditional_load_gate():
assert routing_table() == ROUTES
for names in ROUTES.values():
for name in names:
assert reference_text(name).strip(), f"empty runtime reference: {name}"
def test_core_and_standard_research_fit_the_read_budget():
core = (SKILL_ROOT / "SKILL.md").read_bytes()
assert len(core.splitlines()) <= 500, "root skill exceeds the 500-line read budget"
assert len(core) <= 32_000, "root skill exceeds the 32,000-byte read budget"
routes = routing_table()
assert routes == ROUTES, "classify new routes before measuring the standard path"
standard = {name for condition in STANDARD_CONDITIONS for name in routes[condition]}
assert standard == {"runtime", "research-runbook", "synthesis"}
total = len(core) + sum(len(reference_text(name).encode("utf-8")) for name in standard)
assert total <= 180_000, f"standard configured research path loads {total} bytes"
def test_configured_user_without_global_marker_has_a_bounded_setup_check():
core = root_text()
gate = core.split("**FIRST-RUN GATE", 1)[1].split("**Onboarding consent", 1)[0]
assert 'grep -q "SETUP_COMPLETE=true" ~/.config/last30days/.env' in gate
assert "FIRST_RUN_DETECTED" in gate
assert "read the setup-wizard reference" in gate
setup = reference_text("setup-wizard")
assert "Skip Step 0" in setup.split("## Step 0: First-Run Setup Wizard", 1)[0]
names = {"runtime", "setup-wizard", "research-runbook", "synthesis"}
total = len((SKILL_ROOT / "SKILL.md").read_bytes()) + sum(
len(reference_text(name).encode("utf-8")) for name in names
)
assert total <= 190_000, f"configured path without a global marker loads {total} bytes"
def test_existing_users_can_load_repair_without_restarting_onboarding():
condition = "Source repair requested or indicated by doctor"
assert routing_table().get(condition) == ("setup-wizard",)
assert condition not in STANDARD_CONDITIONS
setup = reference_text("setup-wizard")
entry = setup.split("## Step 0: First-Run Setup Wizard", 1)[0]
assert "including when `SETUP_COMPLETE=true`" in entry
assert "go directly to the applicable **Manual Setup Guide** subsection" in entry
assert "Skip Step 0" in entry
assert "do not write `SETUP_COMPLETE=true`" in entry
assert "do not start research unless it was requested" in entry
assert "explicit browser-cookie consent" in entry
assert "Existing authorization still applies" in entry
assert "a refusal, skip, or no answer is not consent" in entry
assert "only **X on a Grok Bot (repair)**" in entry
assert "never read a browser session or use the Linux / Mac mini repair" in entry
def test_doctor_routes_host_specific_repairs_without_authorizing_them():
runbook = reference_text("research-runbook")
doctor = runbook.split("**Doctor health check:**", 1)[1].split(
"**Grok session expiry handling:**", 1,
)[0]
assert "Source repair requested or indicated by doctor" in doctor
assert "read `references/setup-wizard.md` in full" in doctor
assert "A doctor result is not authorization" in doctor
assert "For a health-check-only request, relay the audit and repair guidance, then stop" in doctor
def test_selected_references_are_read_fully_without_loading_every_mode():
core = root_text()
bootstrap = core.split("## Bootstrap and file reads", 1)[1].split("## Reference routing", 1)[0]
assert "Read only references whose conditions apply" in bootstrap
assert "Read each selected reference in full before its phase" in bootstrap
assert "Do not recursively read every reference" in bootstrap
assert "continue bounded reads until the entire selected reference is loaded" in bootstrap
assert "missing or unreadable" in bootstrap
assert "stop before the affected command/output" in bootstrap
for line in core.splitlines():
if line.lower().startswith(("read all references", "always read every reference")):
pytest.fail(f"unconditional reference loading defeats disclosure: {line}")
def test_paths_stay_bound_to_the_loaded_root_across_hosts():
core = root_text()
bootstrap = core.split("## Bootstrap and file reads", 1)[1].split("## Reference routing", 1)[0]
assert "root SKILL.md actually loaded by the host" in bootstrap
assert "Never bind it to `references/`, the current directory" in bootstrap
assert "Read or equivalent local file capability" in bootstrap
assert "read that replacement root in full, rebind `SKILL_DIR`" in bootstrap
assert "Paths resolve relative to `SKILL_DIR`" in core
metadata = yaml.safe_load(core.split("---", 2)[1])
assert "Read" in metadata["allowed-tools"].split(", ")
assert "references/*" in metadata["metadata"]["openclaw"]["files"]
def test_json_requests_load_the_structured_contract_without_agent_flag():
routes = routing_table()
assert routes["--agent mode or explicit machine-readable JSON"] == ("agent-mode",)
output = root_text().split("## Output contract", 1)[1].split("**BADGE", 1)[0]
assert "Explicit machine-readable JSON uses `--emit=json`" in output
assert "verbatim engine stdout" in output
assert "no prose badge, synthesis, footer, or invitation added" in output
contract = reference_text("agent-mode")
assert "--emit=json" in contract
assert "--json-profile=raw" in contract
@pytest.fixture(scope="module")
def packaged_skill(tmp_path_factory):
checkout = tmp_path_factory.mktemp("skill package source")
skill = checkout / "skills" / "last30days"
shutil.copytree(SKILL_ROOT, skill, ignore=shutil.ignore_patterns("__pycache__", "*.pyc"))
shutil.copy2(ROOT / ".gitattributes", checkout / ".gitattributes")
git_home = checkout / "git-home"
git_home.mkdir()
env = {
**os.environ,
"HOME": str(git_home),
"GIT_CONFIG_GLOBAL": os.devnull,
"GIT_CONFIG_NOSYSTEM": "1",
}
for command in (
["git", "-c", "init.templateDir=", "init", "-q"],
["git", "add", "--", ".gitattributes", "skills/last30days"],
["git", "-c", "user.name=Package test", "-c", "[email protected]",
"-c", "commit.gpgsign=false", "commit", "-qm", "Package current skill files"],
["bash", "skills/last30days/scripts/build-skill.sh"],
):
result = subprocess.run(command, cwd=checkout, env=env, text=True, capture_output=True)
assert result.returncode == 0, result.stdout + result.stderr
with zipfile.ZipFile(checkout / "dist" / "last30days.skill") as archive:
archive.extractall(checkout / "installed skills")
return checkout, checkout / "installed skills" / "last30days", env
def test_skill_archive_contains_the_actual_routed_reference_bytes(packaged_skill):
checkout, installed, env = packaged_skill
assert (installed / "SKILL.md").read_bytes() == (SKILL_ROOT / "SKILL.md").read_bytes()
for names in ROUTES.values():
for name in names:
assert reference_text(name, installed) == reference_text(name)
result = subprocess.run(
["git", "archive", "--format=zip", "HEAD"], cwd=checkout, env=env,
capture_output=True, check=True,
)
with zipfile.ZipFile(io.BytesIO(result.stdout)) as repository_archive:
for name in {name for names in ROUTES.values() for name in names}:
path = f"skills/last30days/references/{name}.md"
assert path in repository_archive.namelist(), f"plugin archive omits {path}"
assert repository_archive.read(path) == (SKILL_ROOT / "references" / f"{name}.md").read_bytes()
def test_packaged_runtime_resolves_from_a_path_with_spaces_outside_cwd(packaged_skill, tmp_path):
_, installed, _ = packaged_skill
assert " " in str(installed)
runtime = reference_text("runtime", installed)
assignment = re.findall(r"^LAST30DAYS_MEMORY_DIR=.*$", runtime, re.M)
assert len(assignment) == 1
result = subprocess.run(
["bash", "-ec", assignment[0] + '\nprintf "%s" "$LAST30DAYS_MEMORY_DIR"'],
cwd=tmp_path,
env={
"PATH": os.defpath,
"HOME": str(tmp_path / "home"),
"SKILL_DIR": str(installed),
"LAST30DAYS_PYTHON": sys.executable,
"LAST30DAYS_CONFIG_DIR": str(tmp_path / "config"),
"LAST30DAYS_CACHE_DIR": str(tmp_path / "cache"),
"LAST30DAYS_MEMORY_DIR": str(tmp_path / "saved research"),
},
text=True, capture_output=True, timeout=15,
)
assert result.returncode == 0, result.stderr
assert result.stdout == str(tmp_path / "saved research")