@the-open-engine/zeroshot 5.4.0

package/cluster-templates/base-templates/heavy-validation.json

@@ -0,0 +1,272 @@
+{
+  "name": "Heavy Validation",
+  "description": "Stage 2: Security + Adversarial testing. Expensive (120-180s). Only runs after Stage 1 passes.",
+  "params": {
+    "validator_level": {
+      "type": "string",
+      "enum": ["level1", "level2", "level3"],
+      "default": "level3"
+    },
+    "max_tokens": {
+      "type": "number",
+      "default": 100000
+    },
+    "timeout": {
+      "type": "number",
+      "default": 0,
+      "description": "Task timeout in milliseconds (0 = no timeout)"
+    }
+  },
+  "agents": [
+    {
+      "id": "validator-security",
+      "role": "validator",
+      "modelLevel": "{{validator_level}}",
+      "timeout": "{{timeout}}",
+      "maxRetries": 3,
+      "outputFormat": "json",
+      "jsonSchema": {
+        "type": "object",
+        "properties": {
+          "approved": {
+            "type": "boolean"
+          },
+          "summary": {
+            "type": "string"
+          },
+          "errors": {
+            "type": "array",
+            "items": {
+              "type": "string"
+            }
+          }
+        },
+        "required": ["approved", "summary"]
+      },
+      "prompt": {
+        "system": "## 🔴 OUTPUT FORMAT (CRITICAL - READ FIRST)\n\nYour output MUST be MINIMAL and STRUCTURED:\n- Output ONLY the required JSON schema fields\n- NO preambles (\"Here is my analysis...\", \"Let me explain...\")\n- NO verbose summaries - be CONCISE (max 100 chars per string field)\n- NO redundant information\n- NO explanations before or after the JSON\n\n## 🚫 YOU ARE A VALIDATOR - READ-ONLY MODE\n\nYour ONLY job is to VALIDATE and OUTPUT JSON.\n- NEVER use Edit, Write, or any file modification tools\n- NEVER fix code - REJECT with errors array instead\n- NEVER create files, modify files, or make commits\n\n## 🚫 YOU CANNOT ASK QUESTIONS\n\nYou are running non-interactively. There is NO USER to answer.\n- NEVER use AskUserQuestion tool\n- NEVER say \"Should I...\" or \"Would you like...\"\n- When unsure: Make the SAFER choice and proceed.\n\n## 🔴 READ CONTEXT FILES FOR REPO-SPECIFIC VALIDATION\n\n**BEFORE approving any implementation:**\n1. Read the repo's context files (README, CONTRIBUTING, Makefile, package.json, etc.)\n2. Look for validation instructions, scripts, or commands the repo specifies\n3. If context files say to run a validation script (validation script, make check, etc.), RUN IT\n4. If the validation script fails, the implementation is NOT complete - REJECT\n\nThis ensures you validate according to THIS repo's standards, not generic rules.\n\n## 🔴 VERIFICATION PROTOCOL (REQUIRED - PREVENTS FALSE CLAIMS)\n\nBefore making ANY claim about security vulnerabilities or missing protections:\n\n1. **SEARCH FIRST** - Use Glob to find ALL relevant files\n2. **READ THE CODE** - Use Read to inspect actual implementation\n3. **GREP FOR PATTERNS** - Use Grep to search for specific code (auth checks, validation, etc.)\n\n**NEVER claim a vulnerability exists without FIRST searching for the relevant code.**\n\nThe worker may have implemented security features in different files than originally planned. If you claim 'missing input validation' without searching, you may miss that validation exists in 'server/middleware/validator.ts' instead of the controller.\n\n### Example Verification Flow:\n```\n1. Claim: 'Missing SQL injection protection'\n2. BEFORE claiming → Grep for 'parameterized', 'prepared', 'escape' in relevant files\n3. BEFORE claiming → Read the actual database query code\n4. ONLY IF NOT FOUND → Add to errors array\n```\n\nYou are a security auditor for a CRITICAL task.\n\n## Security Review Checklist\n1. Input validation (injection attacks)\n2. Authentication/authorization checks\n3. Sensitive data handling\n4. OWASP Top 10 vulnerabilities\n5. Secrets management\n6. Error messages don't leak info\n\n## Output\n- approved: true if no security issues\n- summary: Security assessment\n- errors: Security vulnerabilities found\n\n## 🔴 DEBUGGING METHODOLOGY CHECK\n\nBefore approving, verify the worker didn't take shortcuts:\n\n### Ad Hoc Fix Detection\n- Did worker fix ONE instance? → Grep for similar patterns. If N > 1 exists, REJECT.\n- Example: Fixed null check in `auth.ts:42`? → `grep -r \"similar pattern\" .` - are there others?\n\n### Root Cause vs Symptom\n- Did worker add a workaround? → Find the ACTUAL bug. If workaround hides real issue, REJECT.\n- Example: Added `|| []` fallback? → WHY is it undefined? Fix THAT.\n\n### Lazy Debugging Red Flags (INSTANT REJECT)\n- Worker suggests \"restart the service\" → REJECT (hides the bug)\n- Worker suggests \"clear the cache\" → REJECT (hides the bug)\n- Worker says \"works on my machine\" → REJECT (not a fix)\n- Worker blames the test → REJECT unless they PROVE test is wrong with evidence"
+      },
+      "contextStrategy": {
+        "sources": [
+          {
+            "topic": "ISSUE_OPENED",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "STATE_SNAPSHOT",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "PLAN_READY",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "IMPLEMENTATION_READY",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "QUICK_VALIDATION_PASSED",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          }
+        ],
+        "format": "chronological",
+        "maxTokens": "{{max_tokens}}"
+      },
+      "triggers": [
+        {
+          "topic": "QUICK_VALIDATION_PASSED",
+          "action": "execute_task"
+        }
+      ],
+      "hooks": {
+        "onComplete": {
+          "action": "publish_message",
+          "config": {
+            "topic": "HEAVY_VALIDATION_RESULT",
+            "content": {
+              "text": "{{result.summary}}",
+              "data": {
+                "approved": "{{result.approved}}",
+                "errors": "{{result.errors}}",
+                "validatorId": "validator-security"
+              }
+            }
+          }
+        }
+      }
+    },
+    {
+      "id": "validator-tester",
+      "role": "validator",
+      "modelLevel": "{{validator_level}}",
+      "timeout": "{{timeout}}",
+      "maxRetries": 3,
+      "outputFormat": "json",
+      "jsonSchema": {
+        "type": "object",
+        "properties": {
+          "approved": {
+            "type": "boolean"
+          },
+          "summary": {
+            "type": "string"
+          },
+          "errors": {
+            "type": "array",
+            "items": {
+              "type": "string"
+            }
+          },
+          "testResults": {
+            "type": "string"
+          },
+          "skippedTests": {
+            "type": "array",
+            "description": "Tests that couldn't run due to missing env vars or external service unavailability. Treated as WARNING, not failure.",
+            "items": {
+              "type": "object",
+              "properties": {
+                "test": {
+                  "type": "string",
+                  "description": "Test file or suite name"
+                },
+                "reason": {
+                  "type": "string",
+                  "description": "Why it was skipped (missing env var, unavailable service, etc.)"
+                }
+              },
+              "required": ["test", "reason"]
+            }
+          }
+        },
+        "required": ["approved", "summary"]
+      },
+      "prompt": {
+        "system": "## 🔴 OUTPUT FORMAT (CRITICAL - READ FIRST)\n\nYour output MUST be MINIMAL and STRUCTURED:\n- Output ONLY the required JSON schema fields\n- NO preambles (\"Here is my analysis...\", \"Let me explain...\")\n- NO verbose summaries - be CONCISE (max 100 chars per string field)\n- NO redundant information\n- NO explanations before or after the JSON\n- testResults field: ONLY include pass/fail counts and key errors, NOT full test output\n\n## 🚫 YOU ARE A VALIDATOR - READ-ONLY MODE\n\nYour ONLY job is to VALIDATE and OUTPUT JSON.\n- NEVER use Edit, Write, or any file modification tools\n- NEVER fix code - REJECT with errors array instead\n- NEVER create files, modify files, or make commits\n\n## 🚫 YOU CANNOT ASK QUESTIONS\n\nYou are running non-interactively. There is NO USER to answer.\n- NEVER use AskUserQuestion tool\n- NEVER say \"Should I...\" or \"Would you like...\"\n- When unsure: Make the SAFER choice and proceed.\n\nYou are a TEST EXECUTOR. Your job is to RUN TESTS, not read them.\n\n## 🔴 CORE PRINCIPLE: RUN THE TESTS, DON'T JUST READ THEM\n\n**Reading test code is NOT verification. You must EXECUTE tests.**\n\n- 'Tests look correct' = NOT ACCEPTABLE\n- 'Test output shows 15/15 passing' = ACTUAL VERIFICATION\n\n## 🔴 STEP 1: DISCOVER TEST COMMANDS FROM REPO CONTEXT\n\n1. Read context files (README, CONTRIBUTING, Makefile, package.json, pyproject.toml, Cargo.toml, etc.)\n2. Find the test runner command - DO NOT assume any specific tool\n3. Common patterns: `npm test`, `pytest`, `go test ./...`, `cargo test`, `make test`\n4. Use what the repo specifies, not hardcoded commands\n\n## 🔴 STEP 2: RUN TARGETED TESTS (NOT FULL SUITE)\n\n1. Identify changed files from IMPLEMENTATION_READY message\n2. Find tests RELATED to changed files:\n   - Same directory (e.g., `src/foo.ts` → `src/foo.test.ts`)\n   - Test files that import the changed files\n   - Naming convention matches (e.g., `auth.ts` → `tests/auth.test.ts`)\n3. Run ONLY related tests using repo's \"related tests\" feature if available:\n   - Jest: `--findRelatedTests <files>`\n   - pytest: `pytest <test_files>`\n   - go test: `go test ./path/to/package`\n4. For CRITICAL tasks (auth, payments, security): broader test scope is acceptable\n5. Record output in testResults field\n6. If ANY tests fail → REJECT immediately\n\n**DO NOT run the full test suite when targeted tests suffice.**\n\n## 🔴 STEP 3: RUN REPO-SPECIFIC VALIDATION\n\nIf context files specify validation commands (validation script, make check, etc.):\n1. RUN THEM\n2. Record output\n3. If they fail → REJECT\n\n## FORBIDDEN PATTERNS\n\n- ❌ 'Tests appear to have good coverage' without running them\n- ❌ 'Test assertions look correct' without executing them\n- ❌ 'The test file exists' as evidence of testing\n- ❌ Approving without testResults containing actual test output\n- ❌ Hardcoding `npm test` or any specific test runner without checking repo context\n- ❌ Running full test suite + coverage when only 2 files changed\n\n## 🔴 HANDLING SKIPPED TESTS\n\nIf a test SKIPS due to missing prerequisites (env vars, external services, credentials):\n- Add to skippedTests array with test name and reason\n- Do NOT add to errors array\n- Do NOT reject for skipped tests alone\n- Core tests must still pass\n\nSkipped = WARNING, not FAILURE. Same as CANNOT_VALIDATE in requirements validation.\n\n## APPROVAL CRITERIA\n\nONLY approve if:\n1. You RAN tests related to changed files\n2. All RUNNABLE tests pass (skipped tests = warning only)\n3. Repo-specific validation commands pass\n4. Changed code has test coverage\n\n## Output\n- **approved**: true if runnable tests passed\n- **summary**: Assessment based on test execution\n- **errors**: Issues from running tests\n- **testResults**: Actual test output (REQUIRED)\n- **skippedTests**: Tests that couldn't run (warning only)\n\n## 🔴 DEBUGGING METHODOLOGY CHECK\n\nBefore approving, verify the worker didn't take shortcuts:\n\n### Ad Hoc Fix Detection\n- Did worker fix ONE instance? → Grep for similar patterns. If N > 1 exists, REJECT.\n- Example: Fixed null check in `auth.ts:42`? → `grep -r \"similar pattern\" .` - are there others?\n\n### Root Cause vs Symptom\n- Did worker add a workaround? → Find the ACTUAL bug. If workaround hides real issue, REJECT.\n- Example: Added `|| []` fallback? → WHY is it undefined? Fix THAT.\n\n### Lazy Debugging Red Flags (INSTANT REJECT)\n- Worker suggests \"restart the service\" → REJECT (hides the bug)\n- Worker suggests \"clear the cache\" → REJECT (hides the bug)\n- Worker says \"works on my machine\" → REJECT (not a fix)\n- Worker blames the test → REJECT unless they PROVE test is wrong with evidence"
+      },
+      "contextStrategy": {
+        "sources": [
+          {
+            "topic": "ISSUE_OPENED",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "STATE_SNAPSHOT",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "PLAN_READY",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "IMPLEMENTATION_READY",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "QUICK_VALIDATION_PASSED",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          }
+        ],
+        "format": "chronological",
+        "maxTokens": "{{max_tokens}}"
+      },
+      "triggers": [
+        {
+          "topic": "QUICK_VALIDATION_PASSED",
+          "action": "execute_task"
+        }
+      ],
+      "hooks": {
+        "onComplete": {
+          "action": "publish_message",
+          "config": {
+            "topic": "HEAVY_VALIDATION_RESULT",
+            "content": {
+              "text": "{{result.summary}}",
+              "data": {
+                "approved": "{{result.approved}}",
+                "errors": "{{result.errors}}",
+                "testResults": "{{result.testResults}}",
+                "skippedTests": "{{result.skippedTests}}",
+                "validatorId": "validator-tester"
+              }
+            }
+          }
+        }
+      }
+    },
+    {
+      "id": "consensus-coordinator",
+      "role": "coordinator",
+      "modelLevel": "level1",
+      "timeout": "{{timeout}}",
+      "outputFormat": "json",
+      "jsonSchema": {
+        "type": "object",
+        "properties": {
+          "allApproved": {
+            "type": "boolean"
+          },
+          "summary": {
+            "type": "string"
+          }
+        },
+        "required": ["allApproved", "summary"]
+      },
+      "prompt": {
+        "system": "Check if both validators approved. Output: {\"allApproved\": boolean, \"summary\": \"<50 chars>\"}"
+      },
+      "contextStrategy": {
+        "sources": [
+          {
+            "topic": "HEAVY_VALIDATION_RESULT",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 2
+          }
+        ],
+        "format": "chronological",
+        "maxTokens": "{{max_tokens}}"
+      },
+      "triggers": [
+        {
+          "topic": "HEAVY_VALIDATION_RESULT",
+          "logic": {
+            "engine": "javascript",
+            "script": "const validators = ['validator-security', 'validator-tester']; const stageStart = ledger.findLast({ topic: 'QUICK_VALIDATION_PASSED' }); if (!stageStart) return false; const lastHeavyConsensus = ledger.findLast({ topic: 'VALIDATION_RESULT', since: stageStart.timestamp }); const since = Math.max(stageStart.timestamp, lastHeavyConsensus?.timestamp || 0); const responses = ledger.query({ topic: 'HEAVY_VALIDATION_RESULT', since }); const latestByValidator = new Map(); for (const response of responses) { if (validators.includes(response.sender)) { latestByValidator.set(response.sender, response); } } return validators.every((validatorId) => latestByValidator.has(validatorId));"
+          },
+          "action": "execute_task"
+        }
+      ],
+      "hooks": {
+        "onComplete": {
+          "action": "publish_message",
+          "transform": {
+            "engine": "javascript",
+            "script": "const validationResults = ledger.query({ topic: 'HEAVY_VALIDATION_RESULT' }); return { topic: 'VALIDATION_RESULT', content: { text: result.allApproved ? 'All validations passed' : 'Stage 2 rejected', data: { approved: result.allApproved, stage: 'heavy', summary: result.summary, errors: validationResults.flatMap(r => r.content?.data?.errors || []), qualityGates: validationResults.flatMap(r => r.content?.data?.qualityGates || []) } } };"
+          }
+        }
+      }
+    }
+  ]
+}

package/cluster-templates/base-templates/quick-validation.json

@@ -0,0 +1,285 @@
+{
+  "name": "Quick Validation",
+  "description": "Stage 1: Requirements + Code validation. Fast feedback (30-60s).",
+  "params": {
+    "validator_level": {
+      "type": "string",
+      "enum": ["level1", "level2", "level3"],
+      "default": "level3"
+    },
+    "max_tokens": {
+      "type": "number",
+      "default": 100000
+    },
+    "timeout": {
+      "type": "number",
+      "default": 0,
+      "description": "Task timeout in milliseconds (0 = no timeout)"
+    }
+  },
+  "agents": [
+    {
+      "id": "validator-requirements",
+      "role": "validator",
+      "modelLevel": "{{validator_level}}",
+      "timeout": "{{timeout}}",
+      "maxRetries": 3,
+      "outputFormat": "json",
+      "jsonSchema": {
+        "type": "object",
+        "properties": {
+          "approved": {
+            "type": "boolean"
+          },
+          "summary": {
+            "type": "string"
+          },
+          "errors": {
+            "type": "array",
+            "items": {
+              "type": "string"
+            }
+          },
+          "criteriaResults": {
+            "type": "array",
+            "description": "Status for each acceptance criterion. PASS/FAIL require evidence. CANNOT_VALIDATE requires reason.",
+            "items": {
+              "type": "object",
+              "properties": {
+                "id": {
+                  "type": "string",
+                  "description": "AC1, AC2, etc. from plan"
+                },
+                "status": {
+                  "type": "string",
+                  "enum": ["PASS", "FAIL", "SKIPPED", "CANNOT_VALIDATE"],
+                  "description": "CANNOT_VALIDATE = verification impossible (missing tools, permissions, etc). Treated as PASS with warning."
+                },
+                "evidence": {
+                  "type": "object",
+                  "description": "REQUIRED for PASS/FAIL. Proof of verification - actual command output.",
+                  "properties": {
+                    "command": {
+                      "type": "string"
+                    },
+                    "exitCode": {
+                      "type": "integer"
+                    },
+                    "output": {
+                      "type": "string"
+                    }
+                  }
+                },
+                "reason": {
+                  "type": "string",
+                  "description": "REQUIRED for CANNOT_VALIDATE. WHY verification is impossible (e.g., 'kubectl not installed', 'no SSH access')."
+                }
+              },
+              "required": ["id", "status"]
+            }
+          }
+        },
+        "required": ["approved", "summary"]
+      },
+      "prompt": {
+        "system": "# REQUIREMENTS VALIDATOR\n\n## 🚫 YOU ARE A VALIDATOR - READ-ONLY MODE\n\nYour ONLY job is to VALIDATE and OUTPUT JSON.\n- NEVER use Edit, Write, or any file modification tools\n- NEVER fix code - REJECT with errors array instead\n- NEVER create files, modify files, or make commits\n\nVerify implementation meets ALL requirements from issue. Hold a HIGH BAR.\n\n## WORKFLOW\n1. Read context files (README, CONTRIBUTING, Makefile, package.json, etc.) for repo-specific validation\n2. Parse acceptanceCriteria from PLAN_READY\n3. For EACH criterion: run verification, record evidence\n4. If repo has validation script (validation script, make check, etc.), RUN IT\n\n## VERIFICATION\n- SEARCH before claiming 'missing' (Glob, Grep, Read)\n- RUN commands, capture output as evidence\n- CANNOT_VALIDATE only for: tool not installed, no network, permission denied\n\n## INSTANT REJECT\n- TODO/FIXME/placeholder = REJECT\n- Silent error swallowing = REJECT\n- 'Phase 2 deferred' = REJECT\n- 'Will add tests later' = REJECT\n- ANY priority=MUST criterion fails = REJECT\n\n## APPROVAL\n- approved:true = ALL MUST criteria pass + no blocking issues\n- approved:false = any MUST fails OR incomplete implementation\n\n🚫 NO questions. Make safe choice and proceed.\n\n## 🔴 OUTPUT FORMAT (CRITICAL)\n\nYou MUST return valid JSON with these REQUIRED fields:\n```json\n{\n  \"approved\": boolean,\n  \"summary\": \"<100 chars max>\",\n  \"errors\": [\"blocking issue 1\", \"blocking issue 2\"],\n  \"criteriaResults\": [{\"id\": \"AC1\", \"status\": \"PASS|FAIL|CANNOT_VALIDATE\", \"evidence\": {\"command\": \"...\", \"exitCode\": 0, \"output\": \"<200 chars>\"}, \"reason\": \"for CANNOT_VALIDATE only\"}]\n}\n```\nNo preamble. JSON only."
+      },
+      "contextStrategy": {
+        "sources": [
+          {
+            "topic": "ISSUE_OPENED",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "STATE_SNAPSHOT",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "PLAN_READY",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "IMPLEMENTATION_READY",
+            "priority": "high",
+            "since": "last_agent_start",
+            "strategy": "latest",
+            "amount": 1
+          }
+        ],
+        "format": "chronological",
+        "maxTokens": "{{max_tokens}}"
+      },
+      "triggers": [
+        {
+          "topic": "IMPLEMENTATION_READY",
+          "action": "execute_task"
+        }
+      ],
+      "hooks": {
+        "onComplete": {
+          "action": "publish_message",
+          "config": {
+            "topic": "QUICK_VALIDATION_RESULT",
+            "content": {
+              "text": "{{result.summary}}",
+              "data": {
+                "approved": "{{result.approved}}",
+                "errors": "{{result.errors}}",
+                "criteriaResults": "{{result.criteriaResults}}",
+                "validatorId": "validator-requirements"
+              }
+            }
+          }
+        }
+      }
+    },
+    {
+      "id": "validator-code",
+      "role": "validator",
+      "modelLevel": "{{validator_level}}",
+      "timeout": "{{timeout}}",
+      "maxRetries": 3,
+      "outputFormat": "json",
+      "jsonSchema": {
+        "type": "object",
+        "properties": {
+          "approved": {
+            "type": "boolean"
+          },
+          "summary": {
+            "type": "string"
+          },
+          "errors": {
+            "type": "array",
+            "items": {
+              "type": "string"
+            }
+          }
+        },
+        "required": ["approved", "summary"]
+      },
+      "prompt": {
+        "system": "# CODE VALIDATOR\n\n## 🚫 YOU ARE A VALIDATOR - READ-ONLY MODE\n\nYour ONLY job is to VALIDATE and OUTPUT JSON.\n- NEVER use Edit, Write, or any file modification tools\n- NEVER fix code - REJECT with errors array instead\n- NEVER create files, modify files, or make commits\n\nSenior engineer code review. Catch REAL bugs, not style preferences.\n\n## WORKFLOW\n1. Read context files (README, CONTRIBUTING, Makefile, package.json, etc.) for repo-specific validation\n2. SEARCH before claiming 'missing' (Glob, Grep, Read)\n3. RUN validation scripts if specified\n\n## INSTANT REJECT\n- TODO/FIXME/placeholder = REJECT\n- Silent error swallowing = REJECT\n- Dangerous fallbacks hiding failures = REJECT\n\n## 🔴 GENERALIZATION CHECK (CRITICAL)\nWorker fixed a bug? Verify they fixed ALL instances:\n1. Identify the PATTERN (not just the line)\n2. `grep -rn \"pattern\" .` - search codebase\n3. If N > 1 exists → Did worker fix ALL? If NO → REJECT\n\nExamples: null check in one handler? Check ALL. SQL injection in one query? Check ALL. A fix that leaves identical bugs elsewhere is NOT a fix.\n\n## 🔴 CLEAN DESIGN CHECK (CRITICAL)\nUnless the issue EXPLICITLY requests backwards compatibility, REJECT any of these:\n- Old function/class kept alongside new one (\"deprecated but still works\") → REJECT. Delete old, update ALL callers.\n- Re-exports or wrappers forwarding to new implementation → REJECT. Update imports at call sites.\n- `_unused` parameter renames to preserve old signatures → REJECT. Change the signature, update callers.\n- Fallback paths handling \"old format\" or \"legacy data\" → REJECT. Migrate the data, remove the fallback.\n- Comments like \"kept for backwards compatibility\" → REJECT. Remove the old code.\n- Feature flags toggling between old and new behavior → REJECT. Ship the new behavior. Delete the old.\n\nThe CLEAN solution: delete the old, update all callers, ship only the current implementation. A senior architect demolishes old scaffolding - they don't leave it standing \"just in case\".\n\n## BLOCKING (reject with WHAT/HOW/WHY)\n- Logic/off-by-one bugs\n- Race conditions\n- Security holes (injection, auth bypass)\n- Resource leaks (timers, connections)\n- God functions (>50 lines) - SPLIT\n- DRY violation (same logic 2+ places)\n- Missing error handling\n- Hardcoded values that should be config\n- Backwards compatibility shims or legacy wrappers (see CLEAN DESIGN CHECK)\n\n## NOT BLOCKING (summary only)\n- Style/naming preferences\n- 'Could theoretically...' without proof\n\n🚫 NO questions. Make safe choice and proceed.\n\n## 🔴 OUTPUT FORMAT (CRITICAL)\n\nYou MUST return valid JSON:\n```json\n{\n  \"approved\": boolean,\n  \"summary\": \"<100 chars max>\",\n  \"errors\": [\"WHAT: X. HOW: Y. WHY: Z\"]\n}\n```\nNo preamble. JSON only."
+      },
+      "contextStrategy": {
+        "sources": [
+          {
+            "topic": "ISSUE_OPENED",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "STATE_SNAPSHOT",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "PLAN_READY",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "IMPLEMENTATION_READY",
+            "priority": "high",
+            "since": "last_agent_start",
+            "strategy": "latest",
+            "amount": 1
+          }
+        ],
+        "format": "chronological",
+        "maxTokens": "{{max_tokens}}"
+      },
+      "triggers": [
+        {
+          "topic": "IMPLEMENTATION_READY",
+          "action": "execute_task"
+        }
+      ],
+      "hooks": {
+        "onComplete": {
+          "action": "publish_message",
+          "config": {
+            "topic": "QUICK_VALIDATION_RESULT",
+            "content": {
+              "text": "{{result.summary}}",
+              "data": {
+                "approved": "{{result.approved}}",
+                "errors": "{{result.errors}}",
+                "validatorId": "validator-code"
+              }
+            }
+          }
+        }
+      }
+    },
+    {
+      "id": "consensus-coordinator",
+      "role": "coordinator",
+      "modelLevel": "level1",
+      "timeout": "{{timeout}}",
+      "outputFormat": "json",
+      "jsonSchema": {
+        "type": "object",
+        "properties": {
+          "allApproved": {
+            "type": "boolean"
+          },
+          "summary": {
+            "type": "string"
+          }
+        },
+        "required": ["allApproved", "summary"]
+      },
+      "prompt": {
+        "system": "Check if both validators approved. Output: {\"allApproved\": boolean, \"summary\": \"<50 chars>\"}"
+      },
+      "contextStrategy": {
+        "sources": [
+          {
+            "topic": "QUICK_VALIDATION_RESULT",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 2
+          }
+        ],
+        "format": "chronological",
+        "maxTokens": "{{max_tokens}}"
+      },
+      "triggers": [
+        {
+          "topic": "QUICK_VALIDATION_RESULT",
+          "logic": {
+            "engine": "javascript",
+            "script": "const lastImpl = ledger.findLast({ topic: 'IMPLEMENTATION_READY' }); const since = lastImpl?.timestamp || 0; return helpers.allResponded(['validator-requirements', 'validator-code'], 'QUICK_VALIDATION_RESULT', since);"
+          },
+          "action": "execute_task"
+        }
+      ],
+      "hooks": {
+        "onComplete": {
+          "action": "publish_message",
+          "config": {
+            "topic": "QUICK_VALIDATION_PASSED",
+            "content": {
+              "text": "Stage 1 passed",
+              "data": {}
+            }
+          },
+          "logic": {
+            "engine": "javascript",
+            "script": "if (!result.allApproved) { const validationResults = ledger.query({ topic: 'QUICK_VALIDATION_RESULT' }); return { topic: 'VALIDATION_RESULT', content: { text: 'Stage 1 rejected', data: { approved: false, stage: 'quick', errors: validationResults.flatMap(r => r.content?.data?.errors || []), qualityGates: validationResults.flatMap(r => r.content?.data?.qualityGates || []) } } }; }"
+          }
+        }
+      }
+    }
+  ]
+}

package/cluster-templates/base-templates/single-worker.json

@@ -0,0 +1,71 @@
+{
+  "name": "Single Worker",
+  "description": "One agent executes and completes. For TRIVIAL tasks and simple INQUIRY.",
+  "params": {
+    "worker_level": {
+      "type": "string",
+      "enum": ["level1", "level2", "level3"],
+      "default": "level1"
+    },
+    "max_tokens": {
+      "type": "number",
+      "default": 50000
+    },
+    "timeout": {
+      "type": "number",
+      "default": 0,
+      "description": "Task timeout in milliseconds (0 = no timeout)"
+    },
+    "task_type": {
+      "type": "string",
+      "enum": ["INQUIRY", "TASK", "DEBUG"],
+      "description": "Type of work"
+    }
+  },
+  "agents": [
+    {
+      "id": "worker",
+      "role": "implementation",
+      "modelLevel": "{{worker_level}}",
+      "timeout": "{{timeout}}",
+      "prompt": {
+        "system": "## 🚫 YOU CANNOT ASK QUESTIONS\n\nYou are running non-interactively. There is NO USER to answer.\n- NEVER use AskUserQuestion tool\n- NEVER say \"Should I...\" or \"Would you like...\"\n- When unsure: Make the SAFER choice and proceed.\n\nYou are an agent handling a {{task_type}} task.\n\n## TASK TYPE: {{task_type}}\n\n{{#if task_type == 'INQUIRY'}}\nThis is an INQUIRY - exploration and understanding only.\n- Answer questions about the codebase\n- Explore files and explain how things work\n- DO NOT make any changes\n- Provide clear, accurate information\n{{/if}}\n\n{{#if task_type == 'TASK'}}\nThis is a TRIVIAL TASK - quick execution.\n- Straightforward, well-defined action\n- Quick to complete (< 15 minutes)\n- Low risk of breaking existing functionality\n- Execute efficiently, verify it works, done\n{{/if}}\n\n{{#if task_type == 'DEBUG'}}\nThis is a TRIVIAL DEBUG - simple fix.\n- Obvious issue with clear solution\n- Fix the root cause, not symptoms\n- Verify the fix works\n{{/if}}"
+      },
+      "contextStrategy": {
+        "sources": [
+          {
+            "topic": "ISSUE_OPENED",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          },
+          {
+            "topic": "STATE_SNAPSHOT",
+            "priority": "required",
+            "strategy": "latest",
+            "amount": 1
+          }
+        ],
+        "format": "chronological",
+        "maxTokens": "{{max_tokens}}"
+      },
+      "triggers": [
+        {
+          "topic": "ISSUE_OPENED",
+          "action": "execute_task"
+        }
+      ],
+      "hooks": {
+        "onComplete": {
+          "action": "publish_message",
+          "config": {
+            "topic": "CLUSTER_COMPLETE",
+            "content": {
+              "text": "Task completed."
+            }
+          }
+        }
+      }
+    }
+  ]
+}