richfrem/agent-plugins-skills / plugins/agent-orchestration/.apm/skills/orchestrating-agent-work
orchestrating-agent-work
VerifiedInteractive orchestrator for complex engineering work. Analyzes the task, conducts a diagnostic interview to select the appropriate execution strategy (Directed Acyclic Graph versus Loop, and which Loop type), plans the artifact, and supervises execution and verification. Use for "how should I orchestrate this", "plan this workflow", "should this be a DAG or a loop", "run these jobs in parallel", "delegate this to a worker agent", or "run/resume this graph manifest". Not for single trivial edits or skill self-improvement.
tests/test_loop_strategies.py
View on GitHub#!/usr/bin/env python3
"""
Unit tests for agent-orchestration/ skills, pattern selection decision tree, and contracts.
Validates:
- Decision tree logic for all orchestration patterns (asserts pattern = graph, not skill = graph-planner)
- Skill metadata standards (kebab-case, third-person description, line budgets)
- evals.json schema compliance (should_trigger boolean, prompt string)
- Elimination of dead CLI subcommand references in orchestrator
- Pattern-owned plan artifact assertions
"""
import json
import os
import re
import sys
import unittest
from pathlib import Path
REPO_ROOT = Path(__file__).resolve().parent.parent.parent.parent
PLUGIN_ROOT = Path(__file__).resolve().parent.parent
def evaluate_loop_strategy(task_description: str, has_dependencies: bool = False,
is_batch: bool = False, is_adversarial: bool = False,
is_bounded: bool = False, requires_human_gate: bool = False,
is_readonly_dag: bool = False) -> str:
"""
Deterministic implementation of select-loop-strategy 6-gate decision tree.
Returns the selected PATTERN name ('graph', 'agent-swarm', 'red-team-review', 'dual-loop', etc.).
Note: requires_human_gate is orthogonal governance and does NOT force pattern = 'graph'.
"""
if has_dependencies or is_readonly_dag:
return "graph"
if is_batch:
return "agent-swarm"
if is_adversarial:
return "red-team-review"
if is_bounded:
return "dual-loop"
return "learning-loop"
class TestLoopStrategiesDecisionTree(unittest.TestCase):
def setUp(self):
"""Point at the real agent-orchestration skills/ directory to audit."""
self.skills_dir = PLUGIN_ROOT / "skills"
def test_all_expected_skills_exist(self):
"""Assert all core loop, graph planning, and execution skills exist in the plugin."""
expected_skills = [
"orchestrator",
"select-loop-strategy",
"dual-loop",
"learning-loop",
"co-pilot-loop",
"agent-swarm",
"red-team-review",
"triple-loop-learning",
"graph-planner",
"graph-execution",
]
for skill_name in expected_skills:
skill_path = self.skills_dir / skill_name / "SKILL.md"
self.assertTrue(
skill_path.exists(),
f"Expected skill {skill_name}/SKILL.md does not exist",
)
def test_routing_decision_tree_patterns(self):
"""
Assert decision tree returns pattern names (e.g. pattern = graph),
and verifies that human gates alone do not force graph routing.
"""
# 1. Structural dependencies route to graph pattern
pattern = evaluate_loop_strategy("Multi-phase refactor with parallel reads and ordered mutations", has_dependencies=True)
self.assertEqual(pattern, "graph", "Structural dependencies must select pattern = 'graph'")
# 2. Read-only DAG routes to graph pattern
pattern = evaluate_loop_strategy("Multi-repo dependency scan joining at barrier", is_readonly_dag=True)
self.assertEqual(pattern, "graph", "Read-only DAG must select pattern = 'graph'")
# 3. Bounded bug fix with human review required routes to dual-loop (NOT graph!)
pattern = evaluate_loop_strategy("Fix off-by-one bug with human approval gate", is_bounded=True, requires_human_gate=True)
self.assertEqual(pattern, "dual-loop", "Bounded bug fix with human gate must select pattern = 'dual-loop', not 'graph'")
# 4. Batch parallel tasks route to agent-swarm
pattern = evaluate_loop_strategy("Convert 50 files independently", is_batch=True)
self.assertEqual(pattern, "agent-swarm", "Batch conversion must select pattern = 'agent-swarm'")
# 5. Security audit routes to red-team-review
pattern = evaluate_loop_strategy("Stress-test auth flow against OWASP top 10", is_adversarial=True)
self.assertEqual(pattern, "red-team-review", "Adversarial audit must select pattern = 'red-team-review'")
def test_skill_frontmatter_and_line_budgets(self):
"""Assert all skills have valid YAML frontmatter and line budget <= 300 lines."""
for skill_dir in self.skills_dir.iterdir():
if not skill_dir.is_dir():
continue
skill_md = skill_dir / "SKILL.md"
self.assertTrue(skill_md.exists(), f"Missing SKILL.md in {skill_dir}")
content = skill_md.read_text(encoding="utf-8")
lines = content.splitlines()
# Line budget check: must be <= 300 lines for lean context
self.assertLessEqual(
len(lines), 300,
f"Skill {skill_dir.name} has {len(lines)} lines (exceeds 300 lines)",
)
# Frontmatter check
self.assertTrue(content.startswith("---\n"), f"Missing frontmatter in {skill_dir.name}")
frontmatter_match = re.search(r"^---\n(.*?)\n---", content, re.DOTALL)
self.assertIsNotNone(frontmatter_match, f"Malformed frontmatter in {skill_dir.name}")
fm_text = frontmatter_match.group(1)
# Name match
name_match = re.search(r"^name:\s*([a-z0-9-]+)", fm_text, re.MULTILINE)
self.assertIsNotNone(name_match, f"Missing or invalid name in {skill_dir.name}")
self.assertEqual(name_match.group(1), skill_dir.name, "Skill name does not match directory")
# Description check: third-person (no 'I ' or 'My ')
desc_match = re.search(r"^description:\s*[\"']?(.*?)[\"']?$", fm_text, re.MULTILINE)
self.assertIsNotNone(desc_match, f"Missing description in {skill_dir.name}")
desc_text = desc_match.group(1)
self.assertFalse(desc_text.lower().startswith("i "), f"Description must be third-person in {skill_dir.name}")
def test_evals_schema_compliance(self):
"""Assert every skill with evals.json uses the should_trigger boolean schema."""
for skill_dir in self.skills_dir.iterdir():
if not skill_dir.is_dir():
continue
evals_file = skill_dir / "evals" / "evals.json"
if evals_file.exists():
data = json.loads(evals_file.read_text(encoding="utf-8"))
self.assertIsInstance(data, list, f"{evals_file} must contain a JSON list")
for idx, entry in enumerate(data):
self.assertIn("prompt", entry, f"Entry {idx} missing 'prompt' in {evals_file}")
self.assertIn("should_trigger", entry, f"Entry {idx} missing 'should_trigger' in {evals_file}")
self.assertIsInstance(entry["should_trigger"], bool, f"'should_trigger' must be boolean in {evals_file}")
def test_learning_loop_does_not_contain_triple_loop_typo(self):
"""Verify the copy-paste typo in learning-loop Option B is fixed."""
learning_loop_md = self.skills_dir / "learning-loop" / "SKILL.md"
content = learning_loop_md.read_text(encoding="utf-8")
self.assertNotIn(
"Open the `triple-loop-learning` SKILL",
content,
"learning-loop still contains erroneous cross-reference to triple-loop-learning for Dual Loop option",
)
def test_orchestrator_removes_dead_cli_warnings(self):
"""Verify orchestrator does not advertise unimplemented scan/bundle subcommands."""
orch_md = self.skills_dir / "orchestrator" / "SKILL.md"
content = orch_md.read_text(encoding="utf-8")
self.assertNotIn("A `scan` CLI subcommand is not currently implemented", content)
self.assertNotIn("A `bundle` CLI subcommand is not currently implemented", content)
if __name__ == "__main__":
unittest.main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |