review-work

Review Work

Use the halyard MCP tools to query what has been done — by you, your team, or across the organization. This is useful for standups, catching up after time away, understanding context before starting new work, or reviewing what happened in a previous session.

Quick Start

Search when you have a topic. List when you don't.

// Looking for something specific? Search first.
mcp__plugin_halyard_ask-expert__search_knowledge(query: "authentication refactor")

// General catch-up without a topic? List with filters.
mcp__plugin_halyard_ask-expert__list_knowledge(author: "me", since: "this week")

// Need to see how an entry connects to others? Explore its graph.
mcp__plugin_halyard_ask-expert__explore_knowledge(entry_id: "entry-id")

Available Tools

1. Search Knowledge

Semantic search across all knowledge entries — work summaries, expert Q&A, decisions, and process docs. Use this when you're looking for something specific or want to find relevant context.

mcp__plugin_halyard_ask-expert__search_knowledge(query: "your search query")

Parameters:

Parameter	Description
query	What to search for — uses semantic similarity (required)
type	Filter by type: "WORK_OUTPUT", "DECISION", "PROCESS", "CONTEXT"
author	Filter by author — use "me" for your own entries
since	Time filter: "today", "yesterday", "this week", "7d", "30d", or an ISO date
limit	Max results to return

Examples:

// What decisions were made about the database?
mcp__plugin_halyard_ask-expert__search_knowledge(query: "database", type: "DECISION")

// What did I work on this week?
mcp__plugin_halyard_ask-expert__search_knowledge(query: "work completed", author: "me", since: "this week")

// Find past answers about deployment
mcp__plugin_halyard_ask-expert__search_knowledge(query: "deployment process")

2. List Knowledge

Chronological listing of knowledge entries. Use this when you want to see recent activity without a specific search query — great for standups and catch-ups.

mcp__plugin_halyard_ask-expert__list_knowledge()

Parameters:

Parameter	Description
type	Filter by type: "WORK_OUTPUT", "DECISION", "PROCESS", "CONTEXT"
author	Filter by author — use "me" for your own entries
since	Time filter: "today", "yesterday", "this week", "7d", "30d", or an ISO date
limit	Max results to return (default: 10)
include_content	Include full content in results (default: false)

Examples:

// Everything from today
mcp__plugin_halyard_ask-expert__list_knowledge(since: "today")

// My work output this week
mcp__plugin_halyard_ask-expert__list_knowledge(author: "me", type: "WORK_OUTPUT", since: "this week")

// Recent decisions
mcp__plugin_halyard_ask-expert__list_knowledge(type: "DECISION", limit: 5)

// What happened yesterday?
mcp__plugin_halyard_ask-expert__list_knowledge(since: "yesterday")

3. Explore Knowledge Graph

Explore relationships between knowledge entries — see what supersedes what, evidence chains, and connected entries:

mcp__plugin_halyard_ask-expert__explore_knowledge(
  entry_id: "entry-id"
)

Parameters:

Parameter	Description
entry_id	The knowledge entry ID to explore (required)
depth	How many hops to traverse: 1 or 2 (default: 1)
format	Output format: "list" (default) or "graph" (for visualization)
include_content	Include full content for related entries (default: false)
types	Filter by relation types (e.g., ["DERIVED_FROM", "SUPERSEDES"])

Use this after search_knowledge to understand how an entry connects to related decisions, work, and processes.

4. View User Profile

See a user's expertise areas and recent activity:

mcp__plugin_halyard_ask-expert__get_user_profile()
mcp__plugin_halyard_ask-expert__get_user_profile(since: "this week")
mcp__plugin_halyard_ask-expert__get_user_profile(user_id: "user-id")

Without since, shows accumulated expertise and stats. With since, shows time-scoped activity including conversations, knowledge entries, and sessions.

5. Update User Profile

Update your living profile document with expertise, preferences, or notes:

mcp__plugin_halyard_ask-expert__update_user_profile(
  content: "## Expertise\n- TypeScript/React\n- System design\n\n## Preferences\n- Prefer functional patterns over OOP",
  sections: ["Expertise", "Preferences"]
)

Parameters:

Parameter	Description
content	Full markdown content for the profile (required)
sections	Section names you authored — protected from system rewrites

6. Log Work (When Needed)

If during your review you realize work from the current session should be recorded, use mcp__plugin_halyard_ask-expert__summarize_work:

mcp__plugin_halyard_ask-expert__summarize_work(
  title: "Brief description of what was done",
  summary: "Detailed explanation of the work, context, and decisions made",
  tags: ["relevant", "tags"]
)

Parameters:

Parameter	Description
title	Short title for the work entry (required)
summary	Detailed summary of what was done and why (required)
entry_type	Type: "WORK_OUTPUT" (default), "DECISION", "PROCESS", or "CONTEXT"
tags	Tags for categorization
knowledge_entry_id	ID of an existing entry to update instead of creating new
source_provider	Source system: "github", "slack", "linear", "claude", "codex", "notion"
source_url	Navigable link to the source material (PR, ticket, thread)
source_knowledge_entry_id	ID of another entry this derives from (citation chain)
supersedes_entry_id	ID of an older entry this replaces (marks it outdated)
session_id	Link this entry to a specific agent session

Knowledge Types

Entries in the knowledge base fall into these categories:

Type	What it contains	Created by
WORK_OUTPUT	Code and implementation work summaries	summarize_work
DECISION	Architectural and design decisions with reasoning	summarize_work
PROCESS	How-to documentation and process notes	summarize_work
CONTEXT	Background context, research, or reference material	summarize_work

Note: Expert Q&A entries are created automatically by summarize_conversation and appear in search results, but are not a type filter value you pass to search_knowledge or list_knowledge.

Common Scenarios

Morning standup / catch-up

mcp__plugin_halyard_ask-expert__list_knowledge(author: "me", since: "yesterday")

Starting a new task — find relevant context

mcp__plugin_halyard_ask-expert__search_knowledge(query: "describe the feature or area you're about to work on")

Weekly review

mcp__plugin_halyard_ask-expert__list_knowledge(since: "this week")

Find out what a teammate worked on

mcp__plugin_halyard_ask-expert__search_knowledge(query: "what was done", since: "this week")

Check if a question was already answered

mcp__plugin_halyard_ask-expert__search_knowledge(query: "your question")

Reviewing Reflections

When reviewing past work, look for reflection bullets in work summaries. These capture what worked, what didn't, learnings, and suggestions from previous sessions.

Use reflections to:

• Spot patterns — If multiple sessions mention the same dead end or workaround, that's a signal worth surfacing
• Inform current work — Before starting a task, check if past reflections flagged relevant gotchas or effective approaches
• Surface suggestions — Past sessions may have recommended improvements that haven't been acted on yet

// Find sessions with reflections about a specific area
mcp__plugin_halyard_ask-expert__search_knowledge(query: "reflections on deployment", type: "WORK_OUTPUT")

// Review recent reflections across the team
mcp__plugin_halyard_ask-expert__list_knowledge(type: "WORK_OUTPUT", since: "this week", include_content: true)

When presenting a review to the user, call out any notable reflections — especially recurring themes or unactioned suggestions. Share one that feels relevant and ask if the user has thoughts:

One pattern I noticed across recent sessions: [reflection]. Worth addressing?

Tips

• Search when you have a topic, list when you don't — If the ask is specific ("what do we know about auth?"), use mcp__plugin_halyard_ask-expert__search_knowledge. If it's a general catch-up ("what happened this week?"), use mcp__plugin_halyard_ask-expert__list_knowledge with time/author filters.
• Use time filters — since is your best friend for scoping results to relevant timeframes
• Use "me" for your own work — The author: "me" filter resolves to your user automatically
• Search before asking experts — Someone may have already asked the same question
• Combine filters — Use type + since + author together to get exactly what you need