Geruyang/GAS-Agent-Skills / GAS-Agent-Skills/gas-centralized-development
gas-centralized-development
VerifiedUse when a user requests 集权模式, 集权开发模式技能, 指挥者模式, 中枢开发, or centralized development, or when coupled project work needs unified direction, subordinate supervision, and coordinated integration and release.
| File | Kind | Size |
|---|---|---|
| CHANGELOG.md | markdown | 5.0 KB |
| README.zh-CN.md | markdown | 7.7 KB |
| SKILL.md | skill md | 20.9 KB |
| agents/openai.yaml | config | 330 B |
| evals/runtime-scenarios.json | config | 4.9 KB |
| evals/scenarios.json | config | 41.8 KB |
| references/protocol.md | markdown | 45.3 KB |
| references/research-basis.md | markdown | 5.3 KB |
| references/runtime-evidence.md | markdown | 16.7 KB |
| scripts/gas_runtime.py | script | 22.9 KB |
| scripts/validate_skill.py | script | 23.5 KB |
| templates/command.example.json | config | 1.9 KB |
| templates/decision.example.json | config | 1.1 KB |
| templates/release.example.json | config | 2.0 KB |
| templates/review.example.json | config | 1.5 KB |
| templates/run.example.json | config | 7.1 KB |
| templates/runtime.example.json | config | 4.4 KB |
| templates/supervision.example.json | config | 3.0 KB |
| templates/task.example.json | config | 3.0 KB |
| tests/test_revision_contract.py | script | 3.6 KB |
| tests/test_role_update_contract.py | script | 7.6 KB |
| tests/test_skill.py | script | 6.1 KB |
| tests/test_v3_contract.py | script | 7.7 KB |
tests/test_role_update_contract.py
View on GitHub"""2026-10-04 static/mutation checks; do not prove agent behavior or real readings."""
import importlib.util
import json
import shutil
import tempfile
import unittest
from pathlib import Path
ROOT = Path(__file__).resolve().parents[1]
class RoleUpdateContract(unittest.TestCase):
def setUp(self):
temporary = tempfile.TemporaryDirectory()
self.addCleanup(temporary.cleanup)
self.root = Path(temporary.name) / "skill"
shutil.copytree(ROOT, self.root, ignore=shutil.ignore_patterns("__pycache__"))
def edit(self, name, change):
path = self.root / name
value = json.loads(path.read_text(encoding="utf-8-sig"))
change(value)
path.write_text(json.dumps(value, ensure_ascii=False), encoding="utf-8")
def validate(self):
spec = importlib.util.spec_from_file_location("role_checker", ROOT / "scripts/validate_skill.py")
checker = importlib.util.module_from_spec(spec)
spec.loader.exec_module(checker)
return checker.validate(self.root)
def reject(self, fragment):
result = self.validate()
failures = [c["name"] for c in result["checks"] if not c["ok"]]
self.assertFalse(result["ok"], failures)
self.assertTrue(any(fragment in item for item in failures), failures)
def test_default_valid(self):
self.assertTrue(self.validate()["ok"])
def test_three_executors_configurable(self):
def configure(d):
d["execution_team"].update(executor_count=3, count_source="proposed", planned_distinct_subagents=6)
d["budget"]["max_active_workers"] = 3
self.edit("templates/run.example.json", configure)
self.assertTrue(self.validate()["ok"])
def test_user_count_configurable(self):
self.edit("templates/run.example.json", lambda d: d["execution_team"].update(executor_count=2, count_source="user", planned_distinct_subagents=5))
self.assertTrue(self.validate()["ok"])
def test_count_must_be_positive_integer(self):
for invalid in (0, -1, True, None, "3"):
with self.subTest(invalid=invalid):
self.edit("templates/run.example.json", lambda d: d["execution_team"].update(executor_count=invalid))
self.reject("positive-count")
def test_wrong_team_formula(self):
self.edit("templates/run.example.json", lambda d: d["execution_team"].update(executor_count=3, count_source="user", planned_distinct_subagents=4))
self.reject("team-formula")
def test_multiple_count_not_default(self):
self.edit("templates/run.example.json", lambda d: d["execution_team"].update(executor_count=3, planned_distinct_subagents=6))
self.reject("count-source")
def test_reviewer_cannot_write_code(self):
self.edit("templates/run.example.json", lambda d: d["review_policy"].update(reviewer_may_write_verification_code=True))
self.reject("code-owner")
def test_review_record_cannot_allow_writing(self):
self.edit("templates/review.example.json", lambda d: d.update(reviewer_may_write_verification_code=True))
self.reject("analysis-only")
def test_rerun_existing_is_permitted(self):
doc = json.loads((self.root / "templates/review.example.json").read_text(encoding="utf-8"))
self.assertIn("rerun_existing_commands", doc["allowed_methods"])
self.assertTrue(self.validate()["ok"])
def test_reviewer_input_categories_required(self):
self.edit("templates/run.example.json", lambda d: d["review_policy"].update(primary_inputs=["executor_outputs"]))
self.reject("review-policy:inputs")
def test_task_owner_binding_required(self):
self.edit("templates/task.example.json", lambda d: d["execution_assignment"].update(task_id="other"))
self.reject("development-verification-binding")
def test_no_fabricated_test_owner(self):
self.edit("templates/task.example.json", lambda d: d["test_plan"].update(code_owner_identity="invented"))
self.reject("verification-code-owner-unknown")
def test_monitoring_cannot_claim_observation(self):
self.edit("templates/supervision.example.json", lambda d: d["monitoring"]["board_current"].update(status="PASS", observations=[{"value":0.1}]))
self.reject("board_current:unobserved")
def test_monitoring_cannot_claim_safety(self):
self.edit("templates/supervision.example.json", lambda d: d["monitoring_capabilities"].update(safety_guaranteed=True))
self.reject("no-safety-claim")
def test_monitoring_unknown_threshold_required(self):
self.edit("templates/supervision.example.json", lambda d: d["monitoring"]["board_voltage"].update(threshold_ref="invented"))
self.reject("threshold-unknown")
def test_all_subordinate_status_coverage_required(self):
self.edit("templates/supervision.example.json", lambda d: d["monitoring"].pop("subordinate_agent_status"))
self.reject("monitoring-dimensions")
def test_resource_constraints_required(self):
self.edit("templates/run.example.json", lambda d: d["parallel_dispatch"].update(constraints=["dependencies"]))
self.reject("constraints")
def test_global_serial_strategy_rejected(self):
self.edit("templates/run.example.json", lambda d: d["parallel_dispatch"].update(strategy="always_serial"))
self.reject("parallel-dispatch:strategy")
def test_malformed_additions_rejected_without_crash(self):
for key in ("execution_team", "review_policy", "parallel_dispatch"):
with self.subTest(key=key):
path = self.root / "templates/run.example.json"
original = path.read_bytes()
self.edit("templates/run.example.json", lambda d: d.update({key: None}))
self.reject("object")
path.write_bytes(original)
def test_shared_runtime_owner_declaration_required(self):
self.edit("templates/runtime.example.json", lambda d: d.pop("verification_ownership"))
self.reject("runtime:verification-owner-unknown")
def test_new_scenarios_remain_not_run(self):
doc = json.loads((self.root / "evals/scenarios.json").read_text(encoding="utf-8"))
self.assertGreaterEqual(len(doc["cases"]),58)
self.assertTrue(all(case["status"]=="NOT_RUN" for case in doc["cases"]))
def test_monitoring_freshness_fields_unknown(self):
doc = json.loads((self.root / "templates/supervision.example.json").read_text(encoding="utf-8"))
for dimension in doc["monitoring"].values():
for field in ("sampled_at", "evaluated_at", "freshness_rule_ref", "validity_window"):
self.assertIn(field, dimension)
self.assertIsNone(dimension[field])
def test_freshness_rule_omission_rejected(self):
self.edit("templates/supervision.example.json", lambda d: d["monitoring"]["board_voltage"].pop("freshness_rule_ref",None))
self.reject("board_voltage:freshness-unknown")
def test_fabricated_sample_time_rejected(self):
self.edit("templates/supervision.example.json", lambda d: d["monitoring"]["board_current"].update(sampled_at="2026-10-04T00:00:00Z"))
self.reject("board_current:freshness-unknown")
def test_freshness_and_replacement_cases_not_run(self):
doc = json.loads((self.root / "evals/scenarios.json").read_text(encoding="utf-8"))
tags = {tag for case in doc["cases"] for tag in case.get("tags",[])}
self.assertTrue({"telemetry-freshness", "telemetry-gap-scope", "executor-replacement-fencing"} <= tags)
self.assertTrue(all(case["status"]=="NOT_RUN" for case in doc["cases"]))
if __name__ == "__main__":
unittest.main(verbosity=2)
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |