abi

Synchronize ABI Markdown and docs claims with executable behavior. Use before editing README, AGENTS.md, CLAUDE.md, GEMINI.md, docs/**/*.mdx, walkthroughs, changelogs, or any old ABI Markdown that mentions WDBX, MCP, GPU, clustering, compression, FHE, agents, or benchmarks.

Validate ABI docs and external collateral against repo evidence and docs/contracts/external-claims-audit.mdx. Use when auditing claims, scanning docs for unsupported QPS/sharding/FHE/AES language, checking one claim before publish, or re-confirming the external-claims audit. Agent procedure only — not a CLI or slash command. Pair with abi-doc-claims-sync when editing docs.

Plan and execute ABI repo work from the TODO board, roadmap/spec docs, Zig 0.17 constraints, and validation gates. Use when the user asks to find ABI todos/roadmaps, organize a large implementation from tasks/todo.md, or decide the next safe work slice. Distinct from the goals skill — that skill owns the persistent ledger tasks/goals.md; this skill plans from the active board and north-star specs.

TUI/dashboard superpower. Launch interactive agent REPL, diagnostics dashboard panes, and slash commands via real abi CLI paths. Use when working on abi tui / dashboard / agent tui. There is no /abi-superpower-tui binary.

Build the abi CLI and non-interactively smoke the diagnostics dashboard — render `abi dashboard` one-shot (stdin from /dev/null forces the non-interactive fallback), assert exit 0 and all 5 panels (System, Plugins, WDBX Storage, Scheduler, Memory). Use to smoke-test `abi dashboard`/`abi tui`/`abi --tui` in CI or headless without hanging in the interactive TUI. Complements run-tui (which drives the interactive path via tmux).

Capture, track, and execute goals for the abi Zig project. Use when the user wants to record a goal, see outstanding goals, update goal status, or break a goal into a verified work slice. Distinct from abi-goal-orchestrator (which plans from the TODO board) — this skill owns a persistent goal ledger and drives goal-to-slice execution.

TUI/dashboard OpenCode plugin. Launch interactive agent REPL, diagnostics dashboard panes, and slash commands via real abi CLI paths. There is no /abi-superpower-tui binary.

Set up and operate abi through opencode — how opencode discovers abi skills, slash-commands, and MCP servers. Use when the user asks about opencode config, why an abi skill is not loading in opencode, or how to add a skill to the opencode allowlist.

Build the abi CLI and drive the interactive diagnostics dashboard (`abi dashboard` / `abi tui` / `abi --tui`) through a tmux pty — launch it, screenshot/capture the rendered pane, assert it painted, send the quit key, tear down. Use to run, launch, screenshot, or smoke-test the abi TUI/dashboard headlessly. The one abi surface that needs a real terminal.

Plan abi TUI/dashboard work — the interactive diagnostics dashboard and agent REPL. Use when asked about abi tui / dashboard, pane splits, slash commands, or session save/load. Routes to run-tui, dashboard-smoke, and abi-superpower-tui. A headless fallback exists; tmux is only for the interactive refresh loop.

WDBX compute backend superpower. CPU/GPU/NPU/TPU backend selector with deterministic CPU SIMD fallback. ANE detection, remote TPU dispatch.

Build the abi CLI and report GPU / accelerator / shader / MLIR backend status and the compute-backend matrix (CPU/GPU/NPU/TPU detection + dynamic selection). Use when asked about backend capabilities, why a backend shows accelerated=false, or to capture a hardware/dispatch report.

Build, launch, and drive the abi Zig project — the `abi` CLI and the `abi-mcp` JSON-RPC server. Use when asked to run, start, build, smoke-test, or screenshot abi, drive its CLI subcommands, exercise the MCP server over stdio, or confirm a change works in the real binaries (not just the test suite).

Plan abi SEA (Sparse Evidence Attention) self-learning work — evidence-augmented completion, 8-signal scorer, EMA modulator persistence, and constitution audit. Use when working on ai_learn / complete --learn / evidence recall. Routes to sea-learn-loop, abi-superpower-sea, abi-superpower-constitution, and sea-learning-controller.

Validate the Mintlify docs site (docs/docs.json + docs/**/*.mdx) via `npx mint@latest validate`, plus a stale-.md-reference scan. Local pre-push gate; CI job docs (mint validate) runs the same script on push/PR. Not part of ./build.sh check. User-invocable only.

Sync canonical skills from abi `.agents/skills/` to in-repo CLI skill dirs (`.claude/skills`, `.grok`). Idempotent. Launch with /sync-clis or launch.sh. Distinct from home `~/.grok/scripts/sync-clis.py`.

Validate the Mintlify docs site (docs/docs.json + docs/**/*.mdx) via `npx mint@latest validate`, plus a stale-.md-reference scan. Local pre-push gate; CI job docs (mint validate) runs the same script on push/PR. Not part of ./build.sh check. User-invocable only.

This skill should be used when the user asks to find legacy patterns, technical debt, or modernization targets in ABI — e.g. 'where should we start modernizing this module', 'find outdated patterns in src/', 'audit this file for tech debt' — before any refactor begins.

This skill should be used when the user asks what a modern idiom looks like for error handling, types, or modularity in Zig — e.g. 'show me the modern way to do this', 'what's the idiomatic Zig 0.17 pattern here' — while implementing a refactor.

This skill should be used when the user is ready to execute a refactor plan and asks how to apply it safely — e.g. 'implement this modernization plan', 'extract this safely', 'how do I cut over without breaking things'.

This skill should be used when the user asks how to approach a modernization — e.g. 'should we rewrite or do this incrementally', 'plan a refactor of X', 'what's the risk here' — at the start of significant work.

This skill should be used when the user asks to verify a refactor is done correctly — e.g. 'did I break anything', 'is this refactor complete', 'validate this change meets modern standards' — as the final gate.

This skill should be used when the user asks to find legacy patterns, technical debt, or modernization targets in ABI — e.g. 'where should we start modernizing this module', 'find outdated patterns in src/', 'audit this file for tech debt' — before any refactor begins.

This skill should be used when the user asks what a modern idiom looks like for error handling, types, or modularity in Zig — e.g. 'show me the modern way to do this', 'what's the idiomatic Zig 0.17 pattern here' — while implementing a refactor.

This skill should be used when the user is ready to execute a refactor plan and asks how to apply it safely — e.g. 'implement this modernization plan', 'extract this safely', 'how do I cut over without breaking things'.

This skill should be used when the user asks how to approach a modernization — e.g. 'should we rewrite or do this incrementally', 'plan a refactor of X', 'what's the risk here' — at the start of significant work.

This skill should be used when the user asks to verify a refactor is done correctly — e.g. 'did I break anything', 'is this refactor complete', 'validate this change meets modern standards' — as the final gate.

Set up and operate abi through opencode — how opencode discovers abi skills, slash-commands, and MCP servers. Use when the user asks about opencode config, why an abi skill is not loading in opencode, or how to add a skill to the opencode allowlist.

Manage and maintain AI coding skills via the skill-loop MCP server. Scan, inspect, review, fix, and monitor skill health.

Systematic discovery of legacy patterns, technical debt, and high-value modernization targets in the ABI codebase. Use for clean-slate analysis before refactors.

Concrete before/after modern idioms for errors, types, modularity, and Zig 0.17 patterns in ABI. Use to apply clean modern patterns during refactors.

Safe transformation techniques and implementation playbooks for clean-slate refactors in ABI. Use when executing modernization plans.

Clean-slate planning, strategy selection (direct / phased / parallel), risk assessment for ABI refactors. Use at start of any significant modernization.

Parity, modern quality criteria, and validation checklists for ABI refactors. Use as final step to ensure clean-slate changes meet standards.

Donald Filimon's agent style for ABI and multi-CLI work: terse path-anchored status, claim-honest verification, cursor/ branches with draft PRs, and mid-task skill routing. Use for Donald, /donald-mode, or requests to work in this style.

GPU and Metal backend superpower. Reports backend status, vector ops, hardware capabilities.

GPU and Metal backend superpower. Reports backend status, vector ops, hardware capabilities.

Report current agent/session status and system health — skill-loop run state, scheduler/memory counters, and active context. Use when the user runs /status in the abi REPL or asks what the agent is doing or for system health.

Plan abi AI work — model routing, completion, SEA self-learning, and training. Use when the user asks to plan AI orchestration, choose a model, run a completion, drive the SEA learn loop, or schedule training. Routes to complete-base, sea-learn-loop, agent-plan-train, and the abi-superpower-ai/sea/constitution superpowers.

Plan abi GPU/backend work — Metal on macOS, CPU SIMD fallback, and the disclosed honest-stub backends (accelerator, shaders, mlir, mobile). Use when asked about GPU/backends, why accelerated=false, or when planning backend work. Routes to backend-diagnostics and abi-superpower-gpu. Never claims native CUDA/ANE/Metal-kernel execution — those are disclosed non-goals.