mjmorales

Autonomous task orchestrator with mode dispatch. Absorbs autopilot (plan-exists, skip PRD) and full-auto (no plan, requirements-first) as invocation modes. Auto-scales between simple mode (<=3 steps, sequential, no worktrees) and full mode (4+ steps, parallel worktrees with mandatory principal-architect review). Each run stores state as JSON under .prove/runs/<branch>/<slug>/ (prd.json, plan.json, state.json, reports/*.json) and is mutated only through the run_state CLI. Creates feature branches, runs validation gates, commits per step, and supports rollback via git. Triggers on "orchestrate", "orchestrator", "autopilot", "full auto", "full-auto", "autonomous execution", "run autonomously", "hands-off", "hands-off mode", "implement without me".

Execute a whole scrum milestone or a plan.json task tree as one parallel fan-out run. Triggers on "run the milestone", "execute milestone", "workflow", "fan out the milestone", "parallel milestone", "run the whole task tree", "milestone autopilot". Compiles the dependency graph to a plan, runs its tasks in parallel waves through the orchestrator's full-mode machinery (worktrees, validators, principal-architect review, sequential merge), mirrors task status back to the scrum store, and auto-rebounds on merge conflicts. Raises per-wave fan-out above the orchestrator default.

Clean and compact prove's durable memory layers — team Lore, the Codex (scrum decisions), annotations, and contributor artifacts — keeping the tribal knowledge that grows team accuracy and folding away the rot. Triggers on "janitor", "clean the lore", "compact the lore", "compact the codex", "memory cleanup", "clean up team memory", "prune stale decisions", "tidy tribal knowledge", "lore cleanup". You are the driver: the CLI emits inventories and executes writes; the `memory-janitor` agent judges each entry; a per-team batch gate approves; nothing is ever deleted — consolidation, promotion, and supersession only.

Anchor session context into prove primitives before compaction and rehydrate from them after. Built-in compaction summarizes by recency and drops the claude-prove state an agent needs to reorient; this skill externalizes volatile context into durable anchors (scrum tasks, decisions, run-state, a compact-anchors pointer file) pre-compact, then runs a deterministic reorientation sequence post-compact. Use before a manual /compact, when context is about to auto-compact, or immediately after a compaction. Triggers on "smart compact", "prepare for compaction", "anchor before compact", "context is getting long", "rehydrate", "reorient after compact".

Build or update the content-addressable file index (CAFI) that gives agents routing-hint descriptions for codebase navigation. Triggers on "index the repo", "update the file index", "rebuild the index", "cafi", "file index". You are the driver: the `cafi plan` CLI emits the mechanical delta (walk, triage, hash, diff, batch), you generate the routing-hint descriptions — inline for small deltas, Agent-tool fan-out for medium, a Workflow-tool fan-out for large builds — and land them through the validated `cafi save` gate.

Drive two structured-agent methodologies on prove primitives: the top-down decompose ladder (charter/VISION/milestone → epic → story → task) and AC-gated story-close. Triggers on "decompose", "decompose the milestone", "break this epic into stories", "ladder down", "decompose ladder", "close the story", "story close", "verify acceptance criteria", "AC-gated close", "run the acceptance criteria", "decompose into epics/stories/tasks". You are the driver Claude session: a planning subagent (Agent tool, native structured output) emits each layer's child list, you write children as layered scrum tasks, an AskUserQuestion gate promotes them, and you recurse. Story-close dispatches each acceptance criterion by kind (bash/assert/gate/ agent), writes a verification reasoning-log entry per criterion, promotes the run's durable decisions into the scrum decision store (human-gated), then delegates worktree/validation/review/merge to orchestrator full-mode.

Prune stale cached versions of the prove plugin from Claude Code's plugin cache. Use when superseded versions pile up under plugins/cache and agents read stale skills/references from them, or when reclaiming plugin-cache disk space. Triggers on "clean up cached plugin versions", "prune the plugin cache", "remove old prove versions", "stale plugin cache".

Gather charter, team, or decomposition-kickoff answers through a self-contained HTML intake form instead of a conversational interview, then drive the same writer the conversation would. Triggers on "intake form", "fill out a form", "charter form", "team form", "decompose form", "HTML form", "render an intake form", "form instead of questions". You are the driver: render the form with `claude-prove intake render`, hand the operator the file to fill and copy, read the pasted-back payload, validate it with `claude-prove intake validate`, and map the validated answers onto the existing writer (bootstrap for charter, `scrum team` for team, the decompose ladder for decompose). The form and the interview are two front-ends to ONE writer — never a second writer.

Prune comments to near-zero and rewrite the code to be hyperdescriptive instead — names, types, and tests carry the intent so comments stop poisoning agents. Use to strip comments from a file or directory, drive a tree under the `eir comments` budget, run the comment-removal sweep, or make over-commented host/engine code self-documenting. Triggers on "prune comments", "strip comments", "remove the comments", "make this hyperdescriptive", "comment sweep", "near-zero comments", "decomment", "reduce comments", "sweep this dir for comments".

The funpack runtime model — things, behaviors, signals, commands, and pipelines. Use when designing or reasoning about how a funpack game is structured and scheduled — what a behavior's parameters/return mean, how things communicate via signals, effect closure, the pipeline schedule and tick fold, the slot contracts (Update/Render/Ui/Audio/Startup), and why behaviors are unit-testable. Triggers on "behavior", "signal", "pipeline", "thing", "tick", "spawn/despawn", "effect closure", "render/audio/ui stage", "how does a funpack game work". Also matches gameplay goals phrased without funpack nouns — "how do I make the player move", "the enemy should chase the player", "make the score go up when the player scores", "how do my objects react to each other", "what runs each frame and in what order".

How to lay out and configure a funpack game project — the enforced directory tree, the `.fcfg` config layer (project/entrypoints/builds/tags), directory-derived modules, the schema/seam/behavior module split, derived capabilities, and packages/dependencies. Use when scaffolding a new project, fixing a project-structure error, wiring an entrypoint, registering a @gtag, or adding a dependency. Triggers on "funpack project", "funpack_configs", ".fcfg", "entrypoints", "tags.fcfg", "builds.fcfg", "project layout/structure", "funpack module", "funpack package/dependency", "scaffold a funpack game". Also matches setup goals phrased without funpack terms — "how do I start a new game", "where does this file go", "how do I add a sprite or level or asset", "my game won't build because of its structure", "how do I add a dependency".

Operate the funpack binary on a machine — install, update, roll back, pin, scan, and report status of the funpack toolchain on disk. Use when the user wants to install funpack, update/upgrade to the latest release, check which funpack version is installed or on PATH, pin or switch to a specific version, roll back a bad upgrade, or audit every funpack binary on the machine. Triggers on "install funpack", "update funpack", "upgrade funpack", "what funpack version", "funpack binary status", "scan funpack installs", "pin funpack to <version>", "roll back funpack", "funpack on PATH", "manage the funpack binary". Not for authoring funpack programs, engine/runtime code, or project scaffolding — those are funpack-project / funpack-engine-api / funpack-language.

The funpack determinism contract, fixed-point numerics, the compiler's structural quality gates, @stub/@todo typed holes, and which funpack-mcp tool drives each compile/test/index-query op. Use to understand why a build fails (a gate), how to stay deterministic, how Fixed-point math behaves, how to develop incrementally with typed holes, and which MCP tool to call to compile, test, or query the index. Triggers on "fixed-point", "Fixed/Q32.32", "determinism", "why won't it compile", "quality gate", "effect closure", "exhaustive match", "@stub", "@todo", "--release", "build the project", "run the tests", "compile to check", "query the index".

The funpack `engine.*` standard library — the API surface a game calls. Use when you need a function or type signature — vectors & fixed-point math, the world (View/Spawn/Ref), device-agnostic input, 2D/3D draw commands, audio, UI widgets, navigation, list/map/grid/string/rand, and the model/anim/render3 trio. Triggers on "engine.math", "engine.world", "engine.input", "engine.render", "what function does/returns", "how do I clamp/fold/draw/spawn", "Vec2", "View[T]", "Draw::", "Bindings", "Input.value/axis", "funpack stdlib/API". Also matches capability goals phrased without an API name — "how do I draw a sprite", "how do I read the keyboard or gamepad", "how do I play a sound", "distance between two points", "pick a random number", "clamp a value".

Write and understand funpack `.fun` source — the language's syntax and semantics. Use when authoring, reading, editing, or explaining funpack code — declarations (thing/behavior/signal/pipeline/fn/enum/data/let/test), types, match, the `with` update, lambdas, string interpolation, the @doc/@gtag/@stub/@todo directives, and modules/imports. Triggers on ".fun", "funpack syntax", "write a funpack ...", "how do I declare", "funpack behavior/thing/signal/pipeline", "funpack match/enum/data". Also matches intent-phrased asks that name no funpack term — "how do I write this in funpack", "why won't this compile", "how do I loop over a list", "how do I update one field", "translate this code to funpack".

Triage and resolve the funpack friction log. Use to process ./friction — the dogfood findings (bugs, crashes, missing features, DX-friction) captured while building games. Reconciles each report into the scrum store (find-or-create a tracked task, route it to the owning team) and then drives the durable root-cause fix to done. Triggers on "process the friction", "triage friction", "work the friction log", "fix the friction reports", "reconcile friction with scrum", "bubble up the friction", "friction skill", "go through ./friction".

Conduct a structured 9-question interview at project start, derive a full contract graph from FRAMEWORK.md, and generate the project CLAUDE.md. Run once per project, before any code is written. Triggers on "new project", "start a project", "inception", "scaffold a project".

Audit and clean up source-code comments in any language. Use when asked to audit comments, clean up comments, remove LLM comment smell, fix comment rot, or do a comment cleanup. Removes stale cross-references (Jira/GitHub/RFC/phase refs), temporal/LLM-authorship tells ("for now", "in a real implementation", "previously"), and WHAT-not-WHY narration; keeps WHY/HOW comments.