Three goblins in a trenchcoat pretending to be a senior developer. Claude Code config shaped by ADHD, friction, and spite.
This repository also configures its own agents. See what goblin-mode tells them →
Agent Claude Code
Use this agent to review a proposed feature or solution against design values before implementation. Evaluates sophistication (depth of understanding), empowerment (serves the user), robustness (handles failure gracefully), ethics (manipulation, accessibility, privacy, sustainability) and explainability (can you…
Agent Claude Code
Use this agent when you need to break down a development request into a detailed, actionable implementation plan. Examples: Context: User wants to add a new feature to their project. user: 'I want to add user authentication to my web app' assistant: 'I'll use the implementation-planner agent to create a detailed plan…
Agent Claude Code
Use this agent when switching between projects (Iris, Rhea, Theia, etc.) to quickly rebuild context. Synthesises recent git history, architectural decisions, current work state and key patterns to minimise context-switching friction. Invoke with "What's the state of this project?" or "Catch me up on [project]".
Agent Claude Code
Use this agent to proactively monitor scope during planning and implementation. Detects when plans grow beyond the original ask, when branches touch unrelated files and when step counts signal complexity creep. Fires on checkpoints rather than keywords; intervenes before scope creep becomes entrenched. Invoke with "Is…
Agent Claude Code
Use this agent at the end of a session to capture state for next time. Summarises accomplishments, notes uncommitted work, updates task-tracker status, generates a work record entry and writes a handoff note for the next session-orchestrator run. Invoke with "I'm done for today", "wrap up", or triggered by SessionEnd.
Agent Claude Code
Use this agent at the start of a session to build a work plan. Combines project context, task-tracker state, open branches and roadmap priorities into a ranked list of 2-3 things to work on. Primes the session on selection: sets task-tracker status, checks out the right branch, loads relevant context. Invoke with…
Agent Claude Code
Use this agent before shipping to run a multi-dimensional quality check. Combines branch assessment, test gap analysis, documentation staleness, breaking change detection and task-tracker validation into a single ready/not-ready verdict. Invoke with "Am I ready to ship?", "check this branch", or when conversation…
Agent Claude Code
Use this agent to keep task tracker state consistent with git/codebase state. Adapts to whatever task source the project uses (see docs/reference/task-trackers/). Detects branch checkouts, PR creation and merges, then updates matching tasks accordingly. Invoke with "sync tasks" or use as a subagent from…
Agent Claude Code
Use this agent to identify code that should have tests but doesn't, using risk-based prioritisation from the clod-role-testingobsessive skill. Analyses branch diff against main, applies impact × complexity × change frequency matrix, and produces a prioritised list with test stubs for critical items. Invoke with "What…
Agent Claude Code
The judgement half of the web principles check. Reads the eleven principles in skills/web-conventions/principles.md against a wireframe, a scope document, built pages or captures, and returns findings in the observation store contract plus any amendment it wants to propose. Spawned by the web-check skill in wireframe…
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: