Harden agent relay workflow authoring guidance

README.md

@@ -2,24 +2,24 @@
 
 Skills, slash commands, and a Claude Code plugin for building multi-agent systems with Agent Relay.
 
-Package metadata lives in [prpm.json](prpm.json). The repo currently publishes `agent-workforce-skills` version `1.0.5`.
+Package metadata lives in [prpm.json](prpm.json). The repo currently publishes `agent-workforce-skills` version `1.0.9`.
 
 ## Published Skills
 
 | Skill | Version | Description |
 |-------|---------|-------------|
-| [choosing-swarm-patterns](skills/choosing-swarm-patterns/SKILL.md) | 1.1.2 | Pick the right Agent Relay orchestration pattern across the 10 core swarm patterns plus specialized patterns. |
-| [writing-agent-relay-workflows](skills/writing-agent-relay-workflows/SKILL.md) | 1.6.6 | Build multi-agent workflows with WorkflowBuilder, DAG dependencies, verification gates, channels, and chat-native coordination recipes. |
+| [choosing-swarm-patterns](skills/choosing-swarm-patterns/SKILL.md) | 1.1.3 | Pick the right Agent Relay orchestration pattern across the 10 core swarm patterns plus specialized patterns. |
+| [writing-agent-relay-workflows](skills/writing-agent-relay-workflows/SKILL.md) | 1.6.10 | Build multi-agent workflows with WorkflowBuilder, DAG dependencies, verification gates, mandatory Claude-then-Codex review/fix loops with test hardening, channels, and chat-native coordination recipes. |
 | [setting-up-relayfile](skills/setting-up-relayfile/SKILL.md) | 1.1.0 | Set up Relayfile mounts and writeback for provider files through local filesystem access. |
 | [using-agent-relay](skills/using-agent-relay/SKILL.md) | 1.2.0 | Coordinate agents in real time with Relaycast messaging, channels, threads, reactions, search, and webhooks. |
 | [running-headless-orchestrator](skills/running-headless-orchestrator/SKILL.md) | 1.0.4 | Self-bootstrap Agent Relay infrastructure and manage worker agents without human intervention. |
-| [relay-80-100-workflow](skills/relay-80-100-workflow/SKILL.md) | 1.0.4 | Author workflows that close the 80-to-100 validation gap with repair-aware test, verify, and commit gates. |
+| [relay-80-100-workflow](skills/relay-80-100-workflow/SKILL.md) | 1.0.7 | Author workflows that close the 80-to-100 validation gap with repair-aware test, verify, mandatory Claude-then-Codex review/fix with test hardening, and commit gates. |
 
 ## Slash Commands
 
 | Command | Version | Description |
 |---------|---------|-------------|
-| [/create-workflow](commands/create-workflow.md) | 1.0.0 | Scaffold a model-agnostic Agent Relay workflow using the workflow and swarm-pattern skills. |
+| [/create-workflow](commands/create-workflow.md) | 1.0.2 | Scaffold a model-agnostic Agent Relay workflow using the workflow and swarm-pattern skills, including mandatory Claude-then-Codex review/fix loops with test hardening. |
 | [/spawn](commands/spawn.md) | 1.0.0 | Bootstrap the broker and spawn a worker for `claude`, `codex`, `opencode`, `droid`, `gemini`, or `pi`. |
 
 ## Claude Relay Plugin

commands/create-workflow.md

@@ -19,6 +19,7 @@ Scaffold a new multi-agent workflow that runs on the agent-relay broker. This co
 3. **Author the workflow.** Using the WorkflowBuilder API from the skill:
    - Define each step with explicit inputs and `{{steps.X.output}}` chaining.
    - Add a `verify` gate that fails closed on missing evidence.
+   - Add the mandatory sequential fresh-eyes review/fix loops: Claude reviews the real diff/artifacts, a fixer repairs findings and adds or updates appropriate tests/proofs, Claude reviews the post-fix state again, then Codex repeats the same cycle from scratch over the post-Claude-fix state until the bounded loops have `NO_ISSUES_FOUND` or write `BLOCKED_NO_COMMIT`.
    - Keep prompts model-agnostic — never hardcode a specific model name into a step's instructions.
    - Size steps so a single agent can complete them in one focused pass.
 
@@ -31,11 +32,13 @@ Scaffold a new multi-agent workflow that runs on the agent-relay broker. This co
 ## Output Contract
 
 - The workflow source file (TypeScript or YAML — match the surrounding repo convention).
-- A one-paragraph summary: pattern chosen, agent roster, verify gate.
+- A one-paragraph summary: pattern chosen, agent roster, verify gate, and Claude-then-Codex review/fix loops.
 - Integration checklist (5 bullets max).
 
 ## Constraints
 
 - Model-agnostic prompts only. No "use claude-opus" or "use gpt-5" inside step instructions.
 - Verify gates must check evidence (artifacts, test results, file contents), not self-reports.
+- Review fix steps must harden fixes with appropriate tests, fixtures, assertions, or deterministic proof commands whenever the finding is testable.
+- Final acceptance, commit, PR creation, or handoff must depend on the post-Codex-fix review path, not directly on implementation or the Claude-only review path.
 - Do not invent SDK APIs — if the skill doesn't document it, ask before adding.

prpm.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-workforce-skills",
-  "version": "1.0.6",
+  "version": "1.0.9",
   "description": "Skills for multi-agent coordination - swarm patterns, workflow building, relay usage, and headless orchestration",
   "author": "khaliqgant",
   "organization": "agent-relay",
@@ -28,8 +28,8 @@
     },
     {
       "name": "writing-agent-relay-workflows",
-      "version": "1.6.7",
-      "description": "Use when building multi-agent workflows with the relay broker-sdk - covers conversation-shape vs pipeline-shape coordination, WorkflowBuilder API, DAG step dependencies, agent definitions, output chaining via {{steps.X.output}}, verification gates, evidence-based completion, channels, swarm patterns, chat-native coordination recipes (Q/A, broadcast-ack, peer review, standup, hand-off), error handling, event listeners, step sizing, lead+workers team pattern, and parallel wave planning",
+      "version": "1.6.10",
+      "description": "Use when building multi-agent workflows with the relay broker-sdk - covers conversation-shape vs pipeline-shape coordination, mandatory sequential Claude-then-Codex fresh-eyes review/fix loops with test hardening, WorkflowBuilder API, DAG step dependencies, agent definitions, output chaining via {{steps.X.output}}, verification gates, evidence-based completion, channels, swarm patterns, chat-native coordination recipes (Q/A, broadcast-ack, peer review, standup, hand-off), error handling, event listeners, step sizing, lead+workers team pattern, and parallel wave planning",
       "format": "claude",
       "subtype": "skill",
       "languages": [
@@ -107,8 +107,8 @@
     },
     {
       "name": "relay-80-100-workflow",
-      "version": "1.0.5",
-      "description": "Use when writing agent-relay workflows that must fully validate features end-to-end before merging - covers the 80-to-100 pattern with PGlite in-memory Postgres testing, mock sandbox patterns, test-fix-rerun loops, verify gates, and full lifecycle from implementation through passing tests to commit",
+      "version": "1.0.7",
+      "description": "Use when writing agent-relay workflows that must fully validate features end-to-end before merging - covers the 80-to-100 pattern with mandatory sequential Claude-then-Codex fresh-eyes review/fix loops with test hardening, PGlite in-memory Postgres testing, mock sandbox patterns, test-fix-rerun loops, verify gates, and full lifecycle from implementation through passing tests to commit",
       "format": "claude",
       "subtype": "skill",
       "tags": [
@@ -126,8 +126,8 @@
     },
     {
       "name": "create-workflow",
-      "version": "1.0.0",
-      "description": "Slash command to scaffold a new agent-relay multi-agent workflow using the writing-agent-relay-workflows skill - harness-agnostic, model-agnostic prompts, picks the right swarm pattern, and emits a runnable WorkflowBuilder file with verify gates and integration notes",
+      "version": "1.0.2",
+      "description": "Slash command to scaffold a new agent-relay multi-agent workflow using the writing-agent-relay-workflows skill - harness-agnostic, model-agnostic prompts, picks the right swarm pattern, and emits a runnable WorkflowBuilder file with verify gates, mandatory Claude-then-Codex review/fix loops with test hardening, and integration notes",
       "format": "claude",
       "subtype": "slash-command",
       "tags": [
@@ -166,7 +166,7 @@
       "id": "agent-relay-starter",
       "name": "Agent Relay Starter",
       "description": "Essential skills for building multi-agent systems with Agent Relay - swarm pattern selection, workflow authoring, and trail debugging",
-      "version": "1.0.5",
+      "version": "1.0.7",
       "category": "development",
       "tags": [
         "multi-agent",
@@ -186,7 +186,7 @@
           "packageId": "@agent-relay/writing-agent-relay-workflows",
           "version": "latest",
           "required": true,
-          "reason": "WorkflowBuilder API guide for authoring DAG-based multi-agent workflows"
+          "reason": "WorkflowBuilder API guide for authoring DAG-based multi-agent workflows with mandatory Claude-then-Codex review/fix loops and test hardening"
         },
         {
           "packageId": "@agent-workforce/trail-snippet",
@@ -198,7 +198,7 @@
           "packageId": "@agent-relay/relay-80-100-workflow",
           "version": "latest",
           "required": false,
-          "reason": "E2E validation patterns for workflows that must be production-ready before merging"
+          "reason": "E2E validation patterns for workflows that must be production-ready before merging, including dual review/fix loops with test hardening"
         }
       ]
     }

skills/relay-80-100-workflow/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: relay-80-100-workflow
-description: Use when writing agent-relay workflows that must fully validate features end-to-end before merging. Covers the 80-to-100 pattern - going beyond "code compiles" to "feature works, tested E2E locally." Includes repair-before-failure validation gates, PGlite for in-memory Postgres testing, mock sandbox patterns, test-fix-rerun loops, verify gates after every edit, and the full lifecycle from implementation through passing tests to commit.
+description: Use when writing agent-relay workflows that must fully validate features end-to-end before merging. Covers the 80-to-100 pattern - going beyond "code compiles" to "feature works, tested E2E locally." Includes repair-before-failure validation gates, mandatory sequential Claude-then-Codex fresh-eyes review/fix loops with test hardening, PGlite for in-memory Postgres testing, mock sandbox patterns, test-fix-rerun loops, verify gates after every edit, and the full lifecycle from implementation through passing tests to commit.
 ---
 
 # Writing 80-to-100 Validated Workflows
@@ -62,9 +62,9 @@ For high-stakes implementation workflows, validation should include human-like r
 3. Before external review, the implementer writes a self-reflection artifact under `.workflow-artifacts/<task>/` covering spec coverage, changed files, tests/proofs, repo-rule alignment, and known risks.
 4. A fresh self-review agent reads the actual files, AGENTS.md / CLAUDE.md, recent related work, and local conventions. It writes findings to disk.
 5. The implementer repairs valid findings, then deterministic gates rerun from captured output.
-6. After all squads converge, run two independent final reviewers, typically Claude and Codex. They compare notes and write one merged final review artifact.
-7. Spawn fresh fix agents for final-review findings. Those agents self-reflect, then the final reviewers re-check the post-fix state.
-8. Commit or PR creation is allowed only after final deterministic acceptance and post-fix review are green. Otherwise write a `BLOCKED_NO_COMMIT` artifact with exact evidence.
+6. After all squads converge, run the mandatory sequential fresh-eyes review/fix loops: Claude reviews the final diff and artifacts, a fixer repairs valid findings and adds or updates appropriate tests/proofs, Claude reviews the post-fix state again, then Codex repeats the same cycle from scratch over the post-Claude-fix state.
+7. If either final review still finds issues, run another explicit fix pass or write `BLOCKED_NO_COMMIT` with exact evidence.
+8. Commit or PR creation is allowed only after final deterministic acceptance and post-Codex-fix review are green. Otherwise write a `BLOCKED_NO_COMMIT` artifact with exact evidence.
 
 This keeps "100%" tied to both executable evidence and independent review over the final state.