auditor
01Agent Claude CodeCursor
System audit: checks architecture, finds duplicates, detects circular dependencies. Use when update-task done prints TRIGGERAUDITOR (every 200 tasks).
not rated 1 6mo ago A 38 tokens
original MIT
1Stars on the repository
27Mods indexed here, across every type
6mo agoLast push, which is what freshness is scored on
MITLicence, which decides whether bodies are shown
Agent Claude CodeCursor
Records production history to DevExperience DB. Runs immediately after Reviewer. Fills real development data for Learning loop.
not rated 1 6mo ago A 28 tokens
original MIT
developer
03Agent Claude CodeCursor
Executes micro-tasks: writes code, runs tests, fixes errors. Use when scheduler assigns a task or when implementing a specific task.
not rated 1 6mo ago A 30 tokens
original MIT
dispatcher
04Agent Claude CodeCursor
Dispatcher — ONLY monitors and switches Left↔Right. No Terminal. Left/Right do ALL work.
not rated 1 6mo ago A 23 tokens
original MIT
fix-planner
05Agent Claude CodeCursor
Queue owner. Corrects specific tasks identified by Learning Agent. Targeted corrections only — no proactive scanning. Max 10 tasks per run.
not rated 1 6mo ago A 32 tokens
original MIT
git-ops
06Agent Claude CodeCursor
The ONLY agent that performs git commits. Handles add, commit, push after Reviewer approval.
not rated 1 6mo ago A 22 tokens
original MIT
learning
07Agent Claude CodeCursor
Analyzes DevExperience history every 10 tasks. Identifies specific tasks needing correction, passes explicit task IDs to FixPlanner. Updates developer-addendum only (never overwrites developer.md).
not rated 1 6mo ago A 39 tokens
original MIT
meta-planner
08Agent Claude CodeCursor
Optimizes task DAG, analyzes technical debt, resplits tasks. Use when update-task done prints TRIGGERMETAPLANNER (every 50 tasks).
not rated 1 6mo ago A 39 tokens
original MIT
Agent Claude CodeCursor
Left Orchestrator. Dispatches agents only. NEVER writes code or does reviews directly. Alternates with Right.
not rated 1 6mo ago A 28 tokens
original MIT
Agent Claude CodeCursor
Right Orchestrator. Execution only. Dispatches agents only. NEVER writes code or does reviews directly. Dispatched by Dispatcher after Left returns BATCHDONE.
not rated 1 6mo ago A 37 tokens
copy · 88% MIT
planner-sub
11Agent Claude CodeCursor
Expands one epic into 15–25 subtasks. ONLY Planner invokes this. Left/Right must call planner, never planner-sub.
not rated 1 6mo ago A 31 tokens
original MIT
planner
12Agent Claude CodeCursor
Breaks goals into epics, dispatches Sub-Planner per epic. Owns full planning flow. Orchestrator only launches Planner.
not rated 1 6mo ago A 30 tokens
original MIT
researcher
13Agent Claude CodeCursor
Compresses context: summarizes progress, updates architecture, removes noise. Use when update-task done prints TRIGGERRESEARCHER (every 20 tasks).
not rated 1 6mo ago A 38 tokens
original MIT
reviewer
14Agent Claude CodeCursor
Reviews code for quality, correctness, and architecture compliance. Use after each task completion for spec compliance and code quality check.
not rated 1 6mo ago A 27 tokens
original MIT
terminal
15Agent Claude CodeCursor
Runs terminal commands. The ONLY subagent that executes shell commands. All dreamteam and git commands go through Terminal.
not rated 1 6mo ago A 25 tokens
original MIT
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: