./browse/prompts

# Hook Observability and Session Replay Imported from curated first-party documentation sources. ## What this covers Use this skill when you need to understand what hooks fired, what they emitted, and how a session can be replayed after the fact. ## Use this when - Debugging hook-driven automation - Replaying session events after failures - Adding observability without blocking interactive work ## Expected outcomes - Hook activity becomes inspectable and replayable - Operational state survives beyond a single terminal session - Observability stays useful without overwhelming operators ## Source synthesis - EVOKORE-MCP/docs/VOICE_AND_HOOKS.md (https://github.com/mattmre/EVOKORE-MCP/blob/main/docs/VOICE_AND_HOOKS.md) - EVOKORE-MCP/docs/USE_CASES_AND_WALKTHROUGHS.md (https://github.com/mattmre/EVOKORE-MCP/blob/main/docs/USE_CASES_AND_WALKTHROUGHS.md) ## Dedupe notes Focuses on replay and observability instead of importing the broader voice-sidecar guide verbatim. ## Source excerpts ### EVOKORE-MCP/docs/VOICE_AND_HOOKS.md EVOKORE currently has three separate voice-related systems plus a set of hook and observability utilities. They overlap in operator workflows, but they are not the same runtime. ## The three voice-related systems ### 1. ElevenLabs MCP proxy This is the optional `elevenlabs` child server configured in `mcp.config.json`. What it is: - proxied through the EVOKORE router - exposed as prefixed MCP tools - available to any EVOKORE-connected MCP client when configured successfully What it is for: - text-to-speech and other ElevenLabs MCP operations as tools - routing voice-related actions through the standard EVOKORE proxy/security stack Requirements: - `uvx` available on PATH - `ELEVENLABS_API_KEY` set ### 2. VoiceMode VoiceMode is a separate voice-conversation system for Claude Code. What it is: - registered separately from EVOKORE - not routed through EVOKORE’s stdio server - used for bidirectional voice conversation in Claude Code What it is for: - speaking to Claude and hearing spoken responses - using `OPENAI_API_KEY` and VoiceMode’s own runtime Windows note: - VoiceMode relies on `uvx` being directly available - set `OPENAI_API_KEY` in the shell that launches Claude Code ### 3. VoiceSidecar VoiceSidecar is a standalone WebSocket server implemented in `src/Voice ... ### EVOKORE-MCP/docs/USE_CASES_AND_WALKTHROUGHS.md This guide turns the runtime contracts into practical operator flows. ## Walkthrough 1: Adopt a workflow from the skill library Use this when you want EVOKORE to retrieve process guidance before the model starts acting. ### Goal Find and adopt an existing workflow such as `session-wrap`. ### Steps 1. Ask the client to search skills: ```text Search the MCP for a workflow about session wrap-up and continuity. ``` 2. EVOKORE uses `search_skills` and returns matching skills. 3. Ask for a specific skill: ```text Show me help for the session-wrap skill. ``` 4. EVOKORE uses `get_skill_help` and returns the skill’s internal instructions. 5. For broader task matching, ask: ```text Resolve a workflow for wrapping this session, documenting open risks, and preparing the next handoff. ``` 6. EVOKORE uses `resolve_workflow` and injects the top 1-3 relevant workflows directly into the tool response. ### Why this matters - keeps the model grounded in repo-specific process - reduces prompt drift - makes handoff and governance behavior repeatable ## Walkthrough 2: Use a proxied tool that requires HITL approval Use this when the tool is configured as `require_approval` in `permissions.yml`. ### Goal Allow a protected proxied tool call such as `fs_write_f ...

# Human-in-the-Loop Approval Token Workflow Imported from curated first-party documentation sources. ## What this covers Use this workflow when autonomous execution needs a durable handoff from human review back into the agent loop. ## Use this when - Retrying a blocked tool call after approval - Adding a human checkpoint to sensitive actions - Reducing insecure workarounds around approval flows ## Expected outcomes - Approval becomes a reusable tokenized workflow - Agents resume work without losing execution context - Security controls stay explicit instead of being implied ## Source synthesis - EVOKORE-MCP/docs/V2_MULTI_AGENT_WORKFLOWS.md (https://github.com/mattmre/EVOKORE-MCP/blob/main/docs/V2_MULTI_AGENT_WORKFLOWS.md) - EVOKORE-MCP/docs/AGENT33_IMPROVEMENT_INSTRUCTIONS.md (https://github.com/mattmre/EVOKORE-MCP/blob/main/docs/AGENT33_IMPROVEMENT_INSTRUCTIONS.md) ## Dedupe notes Synthesizes the approval-token concept from the main workflow doc and the Agent33 improvement transfer notes. ## Source excerpts ### EVOKORE-MCP/docs/V2_MULTI_AGENT_WORKFLOWS.md With EVOKORE-MCP v2.0 fully operational, we can now leverage 40+ proxied GitHub and Filesystem tools seamlessly within complex multi-agent workflows. The core architectural advancements include: ## 1. Dynamic Tool Prefixing & Indexing Instead of loading 40+ tools into an LLM's context window statically (which causes massive bloat), EVOKORE's `ProxyManager` boots child servers (like `@modelcontextprotocol/server-github` and `@modelcontextprotocol/server-filesystem`) and dynamically prefixes their tools (`github_create_issue`, `fs_write_file`). This prevents namespace collisions while keeping the tools accessible to native skills. ## 2. Human-in-the-Loop (HITL) Security Interceptor Automated multi-agent workflows involving sensitive endpoints (like GitHub write access or file deletion) are governed by EVOKORE's stateless `_evokore_approval_token` architecture. - When an agent attempts a restricted action, the tool call is intercepted and blocked. - The server returns an error explicitly commanding the agent to prompt the human for approval. - Upon approval, the agent retries the exact tool call with the injected token, securely fulfilling the workflow without severing the conversational context. ## 3. Active Skill Orchestration (Native Harnessing) Unlike v1.0 where skills merel ... ### EVOKORE-MCP/docs/AGENT33_IMPROVEMENT_INSTRUCTIONS.md > **Purpose**: Feed this file into Claude Code CLI when working on the Agent33 repo. It contains patterns, architectures, and capabilities proven in EVOKORE-MCP that Agent33 should adopt. --- ## 1. Multi-Server MCP Aggregation Pattern **What Agent33 lacks**: Agent33's MCP server (Phase 43) is a single-endpoint bridge. It doesn't aggregate multiple child MCP servers behind a unified namespace. **What to build**: A proxy layer that spawns and manages multiple child MCP servers from a single config file, presenting them as one unified tool surface. ### Implementation spec: ``` mcp.config.json { "servers": { "github": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-github"], "env": { "GITHUB_TOKEN": "${GITHUB_TOKEN}" } }, "fs": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-filesystem", "./"] }, "elevenlabs": { "command": "uvx", "args": ["elevenlabs-mcp"], "env": { "ELEVENLABS_API_KEY": "${ELEVENLABS_API_KEY}" } } } } ``` **Key patterns from EVOKORE**: - **Tool name prefixing**: Every proxied tool gets renamed `{serverId}_{originalName}` to prevent namespace collisions (e.g., `github_create_issue`, `fs_read_file`). First-registration-wins for duplicates. - **Environment interpolation**: `${VAR}` syntax in `env` blocks resolved ...

# Improvement Cycle Review Wizard Imported from curated first-party documentation sources. ## What this covers Use this workflow when an operator needs a guided improvement cycle with live status, explicit approvals, and linked artifacts. ## Use this when - Operational review sessions with live progress - Approval-heavy improvement cycles - Coordinating review artifacts with execution steps ## Expected outcomes - Live workflow execution remains visible to operators - Artifacts, approvals, and tool requests stay connected - Improvement loops become easier to audit and rerun ## Source synthesis - AGENT33/docs/phase25-26-live-review-walkthrough.md (https://github.com/mattmre/AGENT33/blob/main/docs/phase25-26-live-review-walkthrough.md) - AGENT33/docs/operator-improvement-cycle-and-jupyter.md (https://github.com/mattmre/AGENT33/blob/main/docs/operator-improvement-cycle-and-jupyter.md) ## Dedupe notes Merges AGENT33 live review walkthrough details with the shorter operator-focused improvement-cycle guide. ## Source excerpts ### AGENT33/docs/phase25-26-live-review-walkthrough.md This guide documents the operator flow introduced by the Phase 25 live workflow transport and the Phase 26/27 improvement-cycle review wizard stack. ## Scope - Live workflow execution with run-scoped graph refresh - WebSocket-first status streaming with authenticated SSE fallback - Improvement-cycle preset creation - Explanation artifact generation for `plan_review` and `diff_review` - Linked review creation, risk assessment, L1/L2 signoff, and final approval - Pending tool-approval triage inside the same workflow domain > Note > This walkthrough reflects the review stack built on `codex/session58-phase26-wizard` and validated in `codex/session60-phase22-docs-validation`. If the related PRs are still open, `main` may not yet expose every surface described here. ## Operator Flow ```mermaid sequenceDiagram participant UI as "Frontend Control Plane" participant WF as "Workflow APIs" participant VIZ as "Visualization APIs" participant EXP as "Explanation APIs" participant REV as "Review APIs" participant HITL as "Tool Approval APIs" UI->>WF: "POST /v1/workflows/{name}/execute (single mode, caller run_id)" UI->>VIZ: "GET /v1/visualizations/workflows/{workflow_id}/graph?run_id=..." UI->>WF: "WS /v1/workflows/{run_id}/ws" alt "WebSocket unav ... ### AGENT33/docs/operator-improvement-cycle-and-jupyter.md This guide covers the merged operator surfaces for: - the Phase 26 improvement-cycle review wizard - the Phase 27 canonical workflow presets - the Phase 38 Docker-backed Jupyter kernel workflow Use it when you want the shortest current path from UI entry point to a real workflow run. ## 1. Improvement-Cycle Wizard The wizard is mounted inside the frontend control plane under the `Workflows` domain. ### Entry path 1. Open the frontend at `http://localhost:3000` 2. Authenticate with a bearer token or API key 3. Open `Advanced Settings` 4. Select the `Workflows` domain 5. Use the `Improvement Cycle Wizard` panel at the top of the page ### What the wizard does The wizard stitches together the backend surfaces that previously had to be called manually: - plan review / diff review generation - review creation and risk assessment - L1 and L2 review submission - tool approval request review and decision capture Reference implementation: - frontend: `frontend/src/features/improvement-cycle/ImprovementCycleWizard.tsx` - tests: `frontend/src/features/improvement-cycle/ImprovementCycleWizard.test.tsx` For the detailed review flow, keep using: - [`phase25-26-live-review-walkthrough.md`](phase25-26-live-review-walkthrough.md) ## 2. Canonical Workflow Presets The `Workflows` doma ...