memory-access

Interactive wizard to install and configure the memory-access MCP server and plugin

Use this agent to coordinate parallel task execution across multiple subagents with contract-enforced lock management, schema-validated handoffs, and deterministic state progression. Examples: <example> Context: User wants to implement a multi-file feature with parallel work streams user: "Implement the new auth system — split the API routes, middleware, and tests across agents" assistant: "I'll use the orchestrator agent to decompose this into lock-safe parallel tasks with validated handoffs." <commentary> Multi-agent parallel work requires lock management, assignment packets, and result validation — exactly what the orchestrator enforces. </commentary> </example> <example> Context: User needs to coordinate agents that must not conflict on shared files user: "Run these three refactoring tasks in parallel but make sure agents don't step on each other" assistant: "I'll use the orchestrator agent to manage lock scopes and prevent resource conflicts." <commentary> Lock-scope enforcement and overlap detection are core orchestrator responsibilities. </commentary> </example> <example> Context: A previous multi-agent run was interrupted and needs recovery user: "The agent swarm crashed mid-run — pick up where we left off" assistant: "I'll use the orchestrator agent to reload the ledger, reconcile locks, and resume from the last valid state." <commentary> Crash recovery with ledger reload and orphan detection is a defined orchestrator protocol. </commentary> </example>

Analysis worker for tasks that examine code without modifying it. Use for code review, architecture analysis, dependency audits, security scanning, or research tasks where the output is a report, not a code change. <example> Context: Orchestrator assigns a code review task user: "Review the auth module for security vulnerabilities" assistant: "I'll use worker-readonly to analyze the code and report findings." <commentary> Code review reads files and reports findings. No file mutation needed. </commentary> </example> <example> Context: Orchestrator assigns a dependency analysis task user: "Audit which modules depend on the legacy API client" assistant: "I'll use worker-readonly to trace dependencies via Grep/Glob." <commentary> Dependency tracing is pure analysis — read and search only. </commentary> </example>

Implementation worker for tasks that create or modify files. Use when a task requires writing code, editing configurations, creating test fixtures, or any file mutation. The subagent directive is injected automatically — this agent handles execution. <example> Context: Orchestrator assigns a task to implement a new module user: "Implement the auth middleware in src/middleware/auth.py" assistant: "I'll use the worker-readwrite agent to implement this within the assigned lock scope." <commentary> File creation and editing require Write/Edit tools. worker-readwrite is the standard implementation worker. </commentary> </example> <example> Context: Orchestrator assigns a refactoring task user: "Refactor the database queries to use parameterized statements" assistant: "I'll use worker-readwrite to edit the existing query files." <commentary> Editing existing files requires Edit tool. This is a readwrite task. </commentary> </example>

Test execution worker for running test suites, validating builds, and verifying acceptance criteria through command execution. Use when the task is to run tests, check build outputs, or validate that prior changes work correctly — not to write new code. <example> Context: Orchestrator needs to validate a completed implementation task user: "Run the auth test suite and report results" assistant: "I'll use worker-test to execute the tests and report pass/fail evidence." <commentary> Test execution needs Bash for running commands and Read/Grep for examining output. No file writing. </commentary> </example> <example> Context: Orchestrator needs integration verification across modules user: "Verify the API endpoints return correct responses after the refactor" assistant: "I'll use worker-test to run integration tests and capture results." <commentary> Integration testing is command execution + result analysis. </commentary> </example>

Interactively create and ingest a knowledge base from crawled URLs, single pages, or Firecrawl JSON folders with real-time progress and detailed statistics

Operate strict orchestrator/subagent workflows with enforceable contracts, lock management, and fresh-context handoff execution. Use when planning or running agent swarms, validating assignment/result packets, preventing lock overlap, reconciling task ledgers, or recovering blocked multi-agent runs.

This skill should be used when the user asks to "store a memory", "remember this", "save this insight", "search memories", "find related insights", "what do I know about", "connect these concepts", "add a relationship", "traverse the knowledge graph", or when working with the memory-access MCP tools. Also activates when storing learnings, debugging knowledge, or building on prior insights.