conexus 3.2.0__tar.gz → 3.2.2__tar.gz

{conexus-3.2.0 → conexus-3.2.2}/.claude-plugin/marketplace.json

@@ -9,13 +9,13 @@
       "name": "nx",
       "source": "./nx",
       "description": "Self-hosted three-tier knowledge management with 17 specialized agents, analytical query pipelines, semantic search, and RDR decision tracking for Claude Code.",
-      "version": "3.2.0"
+      "version": "3.2.2"
     },
     {
       "name": "sn",
       "source": "./sn",
       "description": "Injects Serena and Context7 MCP tool usage guidance into subagents via SubagentStart hook.",
-      "version": "3.2.0"
+      "version": "3.2.2"
     }
   ]
 }

{conexus-3.2.0 → conexus-3.2.2}/CHANGELOG.md

@@ -6,6 +6,22 @@ Versioning follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
 
+## [3.2.2] - 2026-04-07
+
+### Fixed
+- **Plugin audit compliance** — added `nx/.claude-plugin/plugin.json` manifest; fixed 9 agents using non-standard `color` values (only `red`, `blue`, `green`, `yellow`, `purple`, `orange`, `pink`, `cyan` are valid per Claude Code docs).
+- **PermissionRequest hooks** — added `sequential-thinking` to nx MCP auto-approve list (was causing CI failure); explicit matchers in hooks.json reverted to wildcard routing (decision logic stays explicit in shell scripts).
+
+## [3.2.1] - 2026-04-07
+
+### Added
+- **File-path extraction linker** — `generate_rdr_filepath_links()` scans RDR content for source file paths and creates `implements` links to matching catalog code entries. `created_by="filepath_extractor"`. Wired into the indexer alongside the existing heuristic linker.
+
+### Fixed
+- **MCP auto-approve hooks** — replaced wildcard glob patterns with explicit full tool name lists in both nx (28 tools) and sn (27 tools) PermissionRequest hooks.
+- **Agent self-seeding** — 5 analysis/research agents now self-seed T1 scratch with `link-context` when dispatched without a skill, so the auto-linker fires regardless of dispatch path.
+- **Mandatory T3 persistence** — added `<HARD-GATE>` and Stop Criteria enforcement for `store_put` in deep-research-synthesizer, deep-analyst, debugger, architect-planner, and codebase-deep-analyzer.
+
 ## [3.2.0] - 2026-04-06
 
 ### Added

{conexus-3.2.0 → conexus-3.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: conexus
-Version: 3.2.0
+Version: 3.2.2
 Summary: Self-hosted semantic search and knowledge management for LLM-driven development
 Project-URL: Homepage, https://github.com/Hellblazer/nexus
 Project-URL: Repository, https://github.com/Hellblazer/nexus

conexus-3.2.2/nx/.claude-plugin/plugin.json

@@ -0,0 +1,11 @@
+{
+  "name": "nx",
+  "version": "3.2.2",
+  "description": "Self-hosted three-tier knowledge management with specialized agents, analytical query pipelines, semantic search, and RDR decision tracking for Claude Code.",
+  "author": {
+    "name": "Hal Hildebrand",
+    "url": "https://github.com/Hellblazer"
+  },
+  "repository": "https://github.com/Hellblazer/nexus",
+  "license": "AGPL-3.0-or-later"
+}

{conexus-3.2.0 → conexus-3.2.2}/nx/CHANGELOG.md

@@ -6,6 +6,22 @@ Versioning follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
 
+## [3.2.2] - 2026-04-07
+
+### Fixed
+- Added `nx/.claude-plugin/plugin.json` manifest
+- Fixed 9 agents with non-standard color values (amber, teal, mint, gold, coral, emerald, indigo, lime)
+- Added `sequential-thinking` to MCP auto-approve hook
+
+## [3.2.1] - 2026-04-07
+
+### Fixed
+- MCP auto-approve hook uses explicit full tool names (28 tools) instead of wildcard
+- 5 agents self-seed link-context from task prompt when dispatched without a skill
+- Mandatory T3 store_put with HARD-GATE enforcement in 5 analysis/research agents
+- rdr-gate stores critique to T3 on both pass and fail
+- rdr-research seeds link-context before agent dispatch
+
 ## [3.2.0] - 2026-04-06
 
 ### Added

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/architect-planner.md

@@ -54,6 +54,14 @@ Before starting, validate the relay contains all required fields per [RELAY_TEMP
 
 T2 memory context is auto-injected by SessionStart and SubagentStart hooks.
 
+### Link Context (before starting work)
+
+Check T1 scratch for existing `link-context` entries via `mcp__plugin_nx_nexus__scratch(action="list")`. If none tagged `link-context`, seed it yourself:
+1. Extract RDR references, document titles, or topic keywords from your task
+2. Resolve to tumblers: `mcp__plugin_nx_nexus__catalog_search(query="<reference>")`
+3. Seed: `mcp__plugin_nx_nexus__scratch(action="put", content='{"targets": [{"tumbler": "<tumbler>", "link_type": "relates"}], "source_agent": "architect-planner"}', tags="link-context")`
+4. If nothing resolves, skip
+
 You are an expert software architect and strategic planner who adapts to any language and build system. Read CLAUDE.md to identify the project's language, build system, module structure, and architectural patterns before starting design work. You excel at creating comprehensive, adaptive execution plans that are self-correcting and goal-oriented.
 
 **Core Responsibilities:**
@@ -162,6 +170,19 @@ Use to propose architectures using proven technologies.
 5. Store design decisions in Nexus: mcp__plugin_nx_nexus__store_put(content="...", collection="knowledge", title="decision-architect-{topic}", tags="architecture"
 
 
+## T3 Persistence (MANDATORY before returning)
+
+You MUST store your architectural decisions to T3 knowledge BEFORE returning. The auto-linker creates catalog links at storage time — those links are lost if you skip this step.
+
+```
+mcp__plugin_nx_nexus__store_put(
+    content="# Architecture: {topic}\n\n{decisions}",
+    collection="knowledge",
+    title="architecture-{topic}-{date}",
+    tags="architecture,architect-planner,{domain}"
+)
+```
+
 ## Recommended Next Step (MANDATORY output)
 
 Your final output MUST include a clearly labeled next-step recommendation for the caller to dispatch `developer`.
@@ -220,3 +241,9 @@ Provide structured plans with:
 7. Bead IDs for all created tasks
 
 Always include a `## Next Step: plan-auditor` block in your output upon plan completion for the caller to dispatch. Be thorough, be complete, be efficient - deliver plans that are executable machines focused on successful outcomes.
+
+<HARD-GATE>
+BEFORE generating your final response, you MUST call store_put to persist your architectural decisions to T3.
+If you have not yet called mcp__plugin_nx_nexus__store_put in this session, STOP and call it NOW.
+Do NOT return without storing. This is not optional.
+</HARD-GATE>

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/codebase-deep-analyzer.md

@@ -3,7 +3,7 @@ name: codebase-deep-analyzer
 version: "2.0"
 description: Performs comprehensive codebase analysis including architecture patterns, dependencies, and technical debt. Use when onboarding to projects, before major refactoring, or for system-wide understanding.
 model: sonnet
-color: amber
+color: yellow
 effort: medium
 ---
 
@@ -54,6 +54,14 @@ Before starting, validate the relay contains all required fields per [RELAY_TEMP
 
 T2 memory context is auto-injected by SessionStart and SubagentStart hooks. Check `/beads:ready` for unblocked tasks.
 
+### Link Context (before starting work)
+
+Check T1 scratch for existing `link-context` entries via `mcp__plugin_nx_nexus__scratch(action="list")`. If none tagged `link-context`, seed it yourself:
+1. Extract RDR references, document titles, or topic keywords from your task
+2. Resolve to tumblers: `mcp__plugin_nx_nexus__catalog_search(query="<reference>")`
+3. Seed: `mcp__plugin_nx_nexus__scratch(action="put", content='{"targets": [{"tumbler": "<tumbler>", "link_type": "relates"}], "source_agent": "codebase-deep-analyzer"}', tags="link-context")`
+4. If nothing resolves, skip
+
 You are an elite codebase architect and analysis specialist with deep expertise in software archaeology, system comprehension, and technical documentation. Your mission is to perform comprehensive, systematic analysis of codebases using sequential thought processes and parallel task coordination.
 
 **Core Analysis Methodology:**
@@ -160,6 +168,19 @@ When analyzing a codebase, check for `docs/rdr/` directory. If present:
 - RDR documents contain architectural decisions, trade-offs, and research — valuable context for codebase understanding
 
 
+## T3 Persistence (MANDATORY before returning)
+
+You MUST store your analysis findings to T3 knowledge BEFORE returning. The auto-linker creates catalog links at storage time — those links are lost if you skip this step.
+
+```
+mcp__plugin_nx_nexus__store_put(
+    content="# Codebase Analysis: {topic}\n\n{findings}",
+    collection="knowledge",
+    title="analysis-codebase-{topic}-{date}",
+    tags="analysis,codebase-deep-analyzer,{domain}"
+)
+```
+
 ## Recommended Next Step (conditional output)
 
 When your analysis reveals work that needs to be planned (e.g., refactoring, new features, debt remediation), your final output MUST include a next-step recommendation for the caller to dispatch `strategic-planner`. Skip if analysis is informational only.
@@ -226,3 +247,9 @@ Store using these naming conventions:
 - Knowledge base stored in Nexus (via store_put tool) for future reference
 
 You approach each codebase as a complex system requiring systematic exploration, patient investigation, and thoughtful synthesis. Your analysis should be thorough enough to enable confident architectural decisions and technical planning.
+
+<HARD-GATE>
+BEFORE generating your final response, you MUST call store_put to persist your findings to T3.
+If you have not yet called mcp__plugin_nx_nexus__store_put in this session, STOP and call it NOW.
+Do NOT return without storing. This is not optional.
+</HARD-GATE>

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/debugger.md

@@ -54,6 +54,14 @@ Before starting, validate the relay contains all required fields per [RELAY_TEMP
 
 T2 memory context is auto-injected by SessionStart and SubagentStart hooks.
 
+### Link Context (before starting work)
+
+Check T1 scratch for existing `link-context` entries via `mcp__plugin_nx_nexus__scratch(action="list")`. If none tagged `link-context`, seed it yourself:
+1. Extract RDR references, document titles, or topic keywords from your task
+2. Resolve to tumblers: `mcp__plugin_nx_nexus__catalog_search(query="<reference>")`
+3. Seed: `mcp__plugin_nx_nexus__scratch(action="put", content='{"targets": [{"tumbler": "<tumbler>", "link_type": "relates"}], "source_agent": "debugger"}', tags="link-context")`
+4. If nothing resolves, skip
+
 ### Sibling Context (check scratch for predecessor findings)
 
 Before forming hypotheses, check if the developer or other agents left context in scratch:
@@ -154,6 +162,19 @@ Pattern: Form hypothesis -> Use search tool to gather evidence -> Validate with
 
 
 
+## T3 Persistence (MANDATORY before returning)
+
+You MUST store your debugging findings to T3 knowledge BEFORE returning. The auto-linker creates catalog links at storage time — those links are lost if you skip this step.
+
+```
+mcp__plugin_nx_nexus__store_put(
+    content="# Debug: {issue}\n\n{findings}",
+    collection="knowledge",
+    title="debug-{issue}-{date}",
+    tags="debug,debugger,{domain}"
+)
+```
+
 ## Recommended Next Step (MANDATORY output)
 
 Your final output MUST include a clearly labeled next-step recommendation for the caller to dispatch `developer`.
@@ -208,3 +229,9 @@ Store using these naming conventions:
 - Suggest preventive measures to avoid similar issues
 
 You approach each debugging session as a scientific investigation, using evidence-based reasoning to systematically eliminate possibilities until the truth emerges. Your goal is not just to fix the immediate problem, but to understand why it occurred and how to prevent similar issues in the future.
+
+<HARD-GATE>
+BEFORE generating your final response, you MUST call store_put to persist your findings to T3.
+If you have not yet called mcp__plugin_nx_nexus__store_put in this session, STOP and call it NOW.
+Do NOT return without storing. This is not optional.
+</HARD-GATE>

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/deep-analyst.md

@@ -54,6 +54,14 @@ Before starting, validate the relay contains all required fields per [RELAY_TEMP
 
 T2 memory context is auto-injected by SessionStart and SubagentStart hooks.
 
+### Link Context (before starting work)
+
+Check T1 scratch for existing `link-context` entries via `mcp__plugin_nx_nexus__scratch(action="list")`. If none tagged `link-context`, seed it yourself:
+1. Extract RDR references, document titles, or topic keywords from your task
+2. Resolve to tumblers: `mcp__plugin_nx_nexus__catalog_search(query="<reference>")`
+3. Seed: `mcp__plugin_nx_nexus__scratch(action="put", content='{"targets": [{"tumbler": "<tumbler>", "link_type": "relates"}], "source_agent": "deep-analyst"}', tags="link-context")`
+4. If nothing resolves, skip
+
 You are a meticulous systems analyst with exceptional analytical capabilities. You specialize in deep investigation, comprehensive understanding, and clear explanation of complex technical problems and systems.
 
 ## Version 2.0 - Enhanced with Hypothesis Testing and Multi-Angle Analysis
@@ -138,6 +146,19 @@ Incorporate or explicitly refute prior findings in Thought 1. When T3 is empty t
 
 
 
+## T3 Persistence (MANDATORY before returning)
+
+You MUST store your analysis findings to T3 knowledge BEFORE returning. The auto-linker creates catalog links at storage time — those links are lost if you skip this step.
+
+```
+mcp__plugin_nx_nexus__store_put(
+    content="# Analysis: {topic}\n\n{findings}",
+    collection="knowledge",
+    title="analysis-deep-{topic}-{date}",
+    tags="analysis,deep-analyst,{domain}"
+)
+```
+
 ## Recommended Next Step (conditional output)
 
 When your investigation reveals issues requiring planned remediation, your final output MUST include a next-step recommendation for the caller to dispatch `strategic-planner`. Skip if findings are informational only.
@@ -262,3 +283,9 @@ Your analysis integrates with:
 
 You are not just an analyst but a detective, scientist, and advisor rolled into one. Your systematic approach, intellectual honesty, and comprehensive methodology ensure that complex problems are not just understood but mastered, with clear paths forward based on solid evidence and rigorous analysis.
 
+<HARD-GATE>
+BEFORE generating your final response, you MUST call store_put to persist your findings to T3.
+If you have not yet called mcp__plugin_nx_nexus__store_put in this session, STOP and call it NOW.
+Do NOT return without storing. This is not optional.
+</HARD-GATE>
+

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/deep-research-synthesizer.md

@@ -3,7 +3,7 @@ name: deep-research-synthesizer
 version: "2.0"
 description: Conducts comprehensive research across nx knowledge store, memory, web resources, and code repositories. Use when needing multi-source research synthesis or building comprehensive understanding of new technologies.
 model: sonnet
-color: teal
+color: cyan
 effort: medium
 ---
 
@@ -55,6 +55,24 @@ Before starting, validate the relay contains all required fields per [RELAY_TEMP
 
 T2 memory context is auto-injected by SessionStart and SubagentStart hooks.
 
+### Link Context (before starting work)
+
+Check T1 scratch for existing `link-context` entries:
+```
+mcp__plugin_nx_nexus__scratch(action="list")
+```
+If no `link-context` tag is present, seed it yourself:
+
+1. Extract RDR references (`RDR-\d+`), document titles, or topic keywords from your task prompt
+2. Resolve to tumblers: `mcp__plugin_nx_nexus__catalog_search(query="<reference>")`
+3. If found, seed scratch:
+   ```
+   mcp__plugin_nx_nexus__scratch(action="put", content='{"targets": [{"tumbler": "<tumbler>", "link_type": "cites"}], "source_agent": "deep-research-synthesizer"}', tags="link-context")
+   ```
+4. If nothing resolves, skip — the auto-linker handles empty context
+
+This ensures your `store_put` calls create catalog links regardless of how you were dispatched.
+
 ## PDF Processing Protocol
 
 1. **First, check if it is already indexed** by searching nx store for the document
@@ -123,12 +141,27 @@ If your project uses beads for task tracking, consider linking research findings
 
 
 
+## T3 Persistence (MANDATORY before returning)
+
+You MUST store your research findings to T3 knowledge BEFORE returning. Do not defer this to knowledge-tidier — the auto-linker creates catalog links at storage time, and those links are lost if you skip this step.
+
+```
+mcp__plugin_nx_nexus__store_put(
+    content="# Research: {topic}\n\n{findings}",
+    collection="knowledge",
+    title="research-{agent}-{topic}-{date}",
+    tags="research,{domain}"
+)
+```
+
+This is not optional. Store first, then recommend knowledge-tidier for consolidation.
+
 ## Recommended Next Step (MANDATORY output)
 
 Your final output MUST include a clearly labeled next-step recommendation for the caller to dispatch `knowledge-tidier`.
 
 **Condition**: ALWAYS after research completion
-**Rationale**: Research findings must be persisted to nx T3 store
+**Rationale**: Knowledge-tidier consolidates and deduplicates stored findings
 **Mechanism**: You do not have the Agent tool — your caller orchestrates the chain. Include this block at the end of your output:
 
 ```
@@ -280,12 +313,13 @@ Track and report:
 
 ## Stop Criteria
 
-Research is complete when:
+Research is NOT complete until ALL of the following are true:
 - All identified sources have been searched
 - All facts have been cross-validated
 - No unresolved contradictions remain
 - Output has been reviewed and versioned
 - Quality metrics meet thresholds
+- **Findings stored to T3 via `store_put`** — this is a hard gate, not optional
 
 ## Edge Case Handling
 
@@ -295,3 +329,9 @@ Research is complete when:
 - **Technical Complexity**: Break down complex topics into digestible components while maintaining accuracy
 
 You are not just a researcher but a knowledge architect, building lasting value in the user information ecosystem with every investigation. Your work creates compounding returns as each research session enriches the collective knowledge base for future inquiries.
+
+<HARD-GATE>
+BEFORE generating your final response, you MUST call store_put to persist your findings to T3.
+If you have not yet called mcp__plugin_nx_nexus__store_put in this session, STOP and call it NOW.
+Do NOT return without storing. This is not optional.
+</HARD-GATE>

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/knowledge-tidier.md

@@ -3,7 +3,7 @@ name: knowledge-tidier
 version: "2.0"
 description: Reviews and consolidates nx T3 knowledge and T2 memory for accuracy and consistency. Use after major research tasks or when contradicting information is discovered across documents.
 model: haiku
-color: mint
+color: green
 effort: medium
 maxTurns: 20
 ---

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/orchestrator.md

@@ -3,7 +3,7 @@ name: orchestrator
 version: "2.0"
 description: Routes requests to appropriate specialized agents and manages multi-agent pipelines. Use when the task is ambiguous, when coordinating work across multiple agents, or when unsure which agent to invoke.
 model: sonnet
-color: gold
+color: yellow
 effort: medium
 ---
 

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/pdf-chromadb-processor.md

@@ -3,7 +3,7 @@ name: pdf-chromadb-processor
 version: "3.0"
 description: Indexes PDF files into nx T3 store for semantic search by delegating to nx index pdf. Use for any PDF that needs to be extracted and made semantically searchable.
 model: haiku
-color: coral
+color: orange
 effort: low
 maxTurns: 30
 ---

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/plan-enricher.md

@@ -3,7 +3,7 @@ name: plan-enricher
 version: "2.0"
 description: Enriches beads with execution context — file paths, code patterns, constraints, test commands, and (when available) audit findings. Use after plan-audit in RDR planning chain, or standalone for bead enrichment within the same session.
 model: sonnet
-color: emerald
+color: green
 effort: medium
 ---
 

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/strategic-planner.md

@@ -3,7 +3,7 @@ name: strategic-planner
 version: "2.1"
 description: Creates phased TDD-driven implementation plans and decomposes complex work into tracked beads. Use for multi-phase feature planning, dependency management, breaking vague requirements into executable tasks, or iterating on existing plans.
 model: opus
-color: indigo
+color: purple
 effort: high
 ---
 
@@ -192,6 +192,21 @@ Each bead must contain sufficient context for autonomous execution:
 
 
 
+## T3 Persistence (MANDATORY before returning)
+
+You MUST store key architectural decisions to T3 knowledge BEFORE returning. Plans go to T2 memory (per Completion Protocol), but validated decisions belong in T3 so the auto-linker can create catalog links.
+
+```
+mcp__plugin_nx_nexus__store_put(
+    content="# Decision: {topic}\n\n{rationale}",
+    collection="knowledge",
+    title="decision-planner-{topic}-{date}",
+    tags="decision,planning,{domain}"
+)
+```
+
+Skip only if the plan contains no architectural decisions (pure task decomposition).
+
 ## Recommended Next Step (MANDATORY output)
 
 Your final output MUST include a clearly labeled next-step recommendation for the caller to dispatch `plan-auditor`.

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/substantive-critic.md

@@ -3,7 +3,7 @@ name: substantive-critic
 version: "2.0"
 description: Provides deep constructive critique of code, documentation, plans, and designs. Identifies structural flaws, logical inconsistencies, and unvalidated assumptions. Use when reviewing architectural decisions, validating implementations against specifications, or auditing plans before committing.
 model: sonnet
-color: teal
+color: pink
 effort: high
 ---
 

{conexus-3.2.0 → conexus-3.2.2}/nx/agents/test-validator.md

@@ -3,7 +3,7 @@ name: test-validator
 version: "2.0"
 description: Verifies test coverage, runs test suites, and validates test quality for code changes. Use after implementation, before marking work complete, or when test failures need systematic root-cause analysis.
 model: sonnet
-color: lime
+color: pink
 effort: high
 ---
 

conexus-3.2.2/nx/hooks/scripts/auto-approve-nx-mcp.sh

@@ -0,0 +1,51 @@
+#!/bin/bash
+# Auto-approve nexus MCP tools — explicit full tool names, no wildcards.
+set -euo pipefail
+
+INPUT=$(cat)
+TOOL_NAME=$(echo "$INPUT" | python3 -c "import json,sys; print(json.loads(sys.stdin.read()).get('tool_name',''))" 2>/dev/null || echo "")
+
+# Explicit allow list — every nx MCP tool by full name
+case "$TOOL_NAME" in
+  mcp__plugin_nx_nexus__search|\
+  mcp__plugin_nx_nexus__query|\
+  mcp__plugin_nx_nexus__store_put|\
+  mcp__plugin_nx_nexus__store_get|\
+  mcp__plugin_nx_nexus__store_list|\
+  mcp__plugin_nx_nexus__store_delete|\
+  mcp__plugin_nx_nexus__memory_put|\
+  mcp__plugin_nx_nexus__memory_get|\
+  mcp__plugin_nx_nexus__memory_search|\
+  mcp__plugin_nx_nexus__memory_delete|\
+  mcp__plugin_nx_nexus__scratch|\
+  mcp__plugin_nx_nexus__scratch_manage|\
+  mcp__plugin_nx_nexus__collection_list|\
+  mcp__plugin_nx_nexus__collection_info|\
+  mcp__plugin_nx_nexus__collection_verify|\
+  mcp__plugin_nx_nexus__plan_save|\
+  mcp__plugin_nx_nexus__plan_search|\
+  mcp__plugin_nx_nexus__catalog_search|\
+  mcp__plugin_nx_nexus__catalog_show|\
+  mcp__plugin_nx_nexus__catalog_list|\
+  mcp__plugin_nx_nexus__catalog_register|\
+  mcp__plugin_nx_nexus__catalog_update|\
+  mcp__plugin_nx_nexus__catalog_link|\
+  mcp__plugin_nx_nexus__catalog_links|\
+  mcp__plugin_nx_nexus__catalog_unlink|\
+  mcp__plugin_nx_nexus__catalog_link_audit|\
+  mcp__plugin_nx_nexus__catalog_link_bulk|\
+  mcp__plugin_nx_nexus__catalog_link_query|\
+  mcp__plugin_nx_nexus__catalog_resolve|\
+  mcp__plugin_nx_nexus__catalog_stats|\
+  mcp__plugin_nx_sequential-thinking__sequentialthinking)
+    python3 -c "
+import json
+print(json.dumps({
+    'hookSpecificOutput': {
+        'hookEventName': 'PermissionRequest',
+        'decision': {'behavior': 'allow'}
+    }
+}))
+"
+    ;;
+esac

{conexus-3.2.0 → conexus-3.2.2}/nx/skills/rdr-gate/SKILL.md

@@ -117,16 +117,18 @@ If no collections found: "No prior RDRs indexed. Cross-project prior-art search
 ### On Pass
 
 1. Write gate result to T2: mcp__plugin_nx_nexus__memory_put(content="outcome: PASSED\ndate: YYYY-MM-DD\ncritical_count: 0\nsignificant_count: N\nobservation_count: N\nsummary: One-sentence summary", project="{repo}_rdr", title="{id}-gate-latest", ttl="permanent", tags="rdr,gate"
-2. Append gate findings to the RDR's Revision History section
-3. Print: `> Run '/nx:rdr-accept <id>' to accept this RDR.`
+2. Store gate critique to T3: mcp__plugin_nx_nexus__store_put(content="# Gate: RDR NNN\n\n{critique}", collection="knowledge", title="gate-rdr-NNN-{date}", tags="rdr,gate,critique")
+3. Append gate findings to the RDR's Revision History section
+4. Print: `> Run '/nx:rdr-accept <id>' to accept this RDR.`
 
 Status remains **Draft** until the author explicitly accepts via `/nx:rdr-accept`.
 
 ### On Fail
 
 1. Write gate result to T2 (same format, `outcome: "BLOCKED"`)
-2. Display the critique with specific sections to address
-3. Status remains Draft
+2. Store gate critique to T3: mcp__plugin_nx_nexus__store_put(content="# Gate BLOCKED: RDR NNN\n\n{critique}", collection="knowledge", title="gate-rdr-NNN-{date}", tags="rdr,gate,critique,blocked")
+3. Display the critique with specific sections to address
+4. Status remains Draft
 
 ## Relay Template (Use This Format)
 

{conexus-3.2.0 → conexus-3.2.2}/nx/skills/rdr-research/SKILL.md

@@ -80,6 +80,17 @@ When the user asks to *investigate* something (not just record a finding):
 - **External research** ("what embedding models support CCE?"): Dispatch `deep-research-synthesizer` agent, then record the finding
 - **Simple recording**: No agent needed — just write the T2 record and update markdown
 
+### Pre-Dispatch: Seed Link Context
+
+Before dispatching an agent, seed T1 scratch with link targets so the auto-linker can create catalog links when the agent stores findings:
+
+1. Resolve the RDR's tumbler: `mcp__plugin_nx_nexus__catalog_search(query="RDR-NNN", content_type="rdr")`
+2. Write link context to scratch:
+   ```
+   mcp__plugin_nx_nexus__scratch(action="put", content='{"targets": [{"tumbler": "<resolved-tumbler>", "link_type": "cites"}], "source_agent": "rdr-research"}', tag="link-context")
+   ```
+3. If catalog search returns no result, skip seeding (the auto-linker handles empty context gracefully)
+
 ## Relay Template (Use This Format)
 
 When dispatching agents (deep-research-synthesizer or codebase-deep-analyzer) via Agent tool, use this exact structure:

{conexus-3.2.0 → conexus-3.2.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "conexus"
-version = "3.2.0"
+version = "3.2.2"
 description = "Self-hosted semantic search and knowledge management for LLM-driven development"
 readme = { file = "README.md", content-type = "text/markdown" }
 requires-python = ">=3.12,<3.14"

{conexus-3.2.0 → conexus-3.2.2}/sn/.claude-plugin/plugin.json

@@ -1,5 +1,5 @@
 {
   "name": "sn",
-  "version": "3.2.0",
+  "version": "3.2.2",
   "description": "Injects Serena and Context7 MCP tool usage guidance into subagents via SubagentStart hook."
 }

conexus-3.2.2/sn/hooks/scripts/auto-approve-sn-mcp.sh

@@ -0,0 +1,46 @@
+#!/bin/bash
+# Auto-approve sn plugin MCP tools (Serena + Context7) — explicit full tool names.
+set -euo pipefail
+
+INPUT=$(cat)
+TOOL_NAME=$(echo "$INPUT" | python3 -c "import json,sys; print(json.loads(sys.stdin.read()).get('tool_name',''))" 2>/dev/null || echo "")
+
+case "$TOOL_NAME" in
+  mcp__plugin_sn_serena__check_onboarding_performed|\
+  mcp__plugin_sn_serena__delete_memory|\
+  mcp__plugin_sn_serena__edit_memory|\
+  mcp__plugin_sn_serena__find_file|\
+  mcp__plugin_sn_serena__initial_instructions|\
+  mcp__plugin_sn_serena__insert_after_symbol|\
+  mcp__plugin_sn_serena__insert_before_symbol|\
+  mcp__plugin_sn_serena__jet_brains_find_declaration|\
+  mcp__plugin_sn_serena__jet_brains_find_implementations|\
+  mcp__plugin_sn_serena__jet_brains_find_referencing_symbols|\
+  mcp__plugin_sn_serena__jet_brains_find_symbol|\
+  mcp__plugin_sn_serena__jet_brains_get_symbols_overview|\
+  mcp__plugin_sn_serena__jet_brains_inline_symbol|\
+  mcp__plugin_sn_serena__jet_brains_move|\
+  mcp__plugin_sn_serena__jet_brains_rename|\
+  mcp__plugin_sn_serena__jet_brains_safe_delete|\
+  mcp__plugin_sn_serena__jet_brains_type_hierarchy|\
+  mcp__plugin_sn_serena__list_dir|\
+  mcp__plugin_sn_serena__list_memories|\
+  mcp__plugin_sn_serena__onboarding|\
+  mcp__plugin_sn_serena__read_memory|\
+  mcp__plugin_sn_serena__rename_memory|\
+  mcp__plugin_sn_serena__replace_symbol_body|\
+  mcp__plugin_sn_serena__search_for_pattern|\
+  mcp__plugin_sn_serena__write_memory|\
+  mcp__plugin_sn_context7__resolve-library-id|\
+  mcp__plugin_sn_context7__query-docs)
+    python3 -c "
+import json
+print(json.dumps({
+    'hookSpecificOutput': {
+        'hookEventName': 'PermissionRequest',
+        'decision': {'behavior': 'allow'}
+    }
+}))
+"
+    ;;
+esac