mikekelly

mikekelly/gap

Settings file Claude Code

Agent settings configuring enabledPlugins.

mikekelly/gap

MCP server Claude CodeCodexCursor +2

Gives the agent up-to-date documentation and code examples for libraries and frameworks, fetched on demand so answers are not stuck at the model's training cutoff. Runs locally from the @upstash/context7-mcp npm package.

mikekelly/gap

MCP server Claude CodeCodexCursor +2

Connect AI agents to Exa for web search, content fetching, and multi-step research. Runs locally from the exa-mcp-server npm package. Needs 1 environment variable to run.

mikekelly/gap

MCP server Claude CodeCodexCursor +2

MCP server "grepapp", hosted remotely at mcp.grep.app, as configured in mikekelly/gap.

mikekelly/gap

Instructions file

Instructions for mikekelly/gap: If your system prompt does NOT mention a Task tool - you're a subagent and this file does NOT apply to you.

mikekelly/promode

Plugin Claude Code

Plugin marketplace listing 1 plugin: promode.

mikekelly/promode

Settings file Claude Code

Agent settings configuring enabledPlugins, worktree, autoMemoryEnabled, autoMemoryDirectory.

mikekelly/promode

Instructions file CodexOpenCode

Instructions for mikekelly/promode, a project described as: A Claude Code plugin designed to make efficient use of Fable on usage credits.

mikekelly/promode

Instructions file

Instructions for mikekelly/promode, covering what is promode?, how the methodology reaches agents, where the brief + hook live; keeping principles in sync and runbooks.

mikekelly/promode

Plugin Claude Code

Promode enhances Claude Code so it builds software more intelligently.

mikekelly/promode

Agent

The evidence side of after-action reviews: analyzes Claude Code agent transcripts to establish what actually happened — checking a subject agent's self-debrief testimony against the record, autopsying dead/stalled/oversized runs, and clustering recurring failure patterns across sessions. Also dispatched to recover a…

mikekelly/promode

Agent

Runs the promode methodology audit end-to-end: fans out parallel read-only assessors (one per alignment dimension) and synthesises their findings into a prioritised, actionable improvement plan for the main agent to ratify. Dispatch when the user wants to assess promode alignment/fit, audit a repo against the…

mikekelly/promode

Agent

Cheap bulk generic executor (Haiku) for simple, well-specified non-code tasks — routine gathering, formatting non-source artifacts, file operations, running existing scripts. NOT for production code changes — those ride TDD in the engineer defs. Haiku has no effort control, so carries no effort field.

mikekelly/promode

Agent

Plans and decides the crucial, hard-to-reverse marketing work: positioning and messaging strategy, brand/category strategy, the channel portfolio, growth strategy, launch strategy, the marketing-plan artifact, and pricing presentation/packaging (pricing level belongs to chief-product-officer /…

mikekelly/promode

Agent

Plans and decides the crucial, hard-to-reverse product work: the goal hierarchy, establishing or majorly revising a persona, kill/build calls, and the evidence backing what users actually need. Marketing one-way doors — positioning, growth, channel strategy — belong to chief-marketing-officer. Drafts recommendations…

mikekelly/promode

Agent

Plans and designs the crucial, hard-to-reverse technical work: software architecture, entity/domain model selection, large-refactor strategy, technology selection, and product calls whose trade-offs are primarily technical (crucial product one-way doors belong to chief-product-officer). Drafts designs and…

mikekelly/promode

Agent

Reviews implementation work. Marks tasks done or requests rework. Pinned to Opus / high effort — review is a judgement seat, not mechanical execution; pass model: sonnet for simple mechanical diffs.

mikekelly/promode

Agent

Walks the knowledge graph (rooted at CLAUDE.md) and ensures every crucial design constraint — invariants, prohibitions, required patterns, load-bearing decisions — is guaranteed-loaded into the nearest loaded CLAUDE.md orientation that governs the affected area. Dispatch when constraints are buried in ADRs, knowledge…

mikekelly/promode

Agent

Investigates failures, analyzes logs, and finds root causes. Produces a reproduction test and reports findings. Does NOT implement fixes unless explicitly asked. Use for debugging, logging analysis, and error investigation. Defaults to Opus (deep-reasoning tier); pass model: sonnet for simple bugs.

mikekelly/promode

Agent

Generic executor at the orchestrator's own tier (inherits the session model): the hardest non-code execution — judge/synthesis-grade grunt-work, deep research, gathering, non-source artifact assembly, running existing scripts. NOT for production code changes — those ride TDD in the engineer defs. Carries no effort…

mikekelly/promode

Agent

Manages dev environments: health checks, docker, services, scripts. Ensures environments are healthy and easily manageable, and commits script/config changes before reporting.

mikekelly/promode

Agent

General-purpose generic executor (Sonnet, medium effort) for routine non-code tasks — research, gathering, formatting non-source artifacts, file operations, doc assembly, running existing scripts. NOT for production code changes — those ride TDD in the engineer defs.

mikekelly/promode

Agent

Drives browsers and GUIs with selector-based discipline: traversals, form-driving, visual checks, and exploratory driving that leave behind deterministic map/graph/recognizer artifacts. Does not change production code — flags code-lane work for engineer re-dispatch. Pinned to Sonnet.

mikekelly/promode

Agent

Deep-reasoning generic executor (Opus, high effort) for non-code tasks that need strong judgement — complex research, synthesis, gathering, non-source artifact assembly, running existing scripts. NOT for production code changes — those ride TDD in the engineer defs.