donaldfilimon/abi / .agents/skills/ai-plan
ai-plan
VerifiedPlan 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.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 1.9 KB |
SKILL.md
View on GitHub---
name: ai-plan
description: 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.
---
# ai-plan
Entry point for abi's AI subsystem (`crates/abi-ai/src/`). Routes to the
specialist skills instead of duplicating them:
| You want to… | Use |
| --- | --- |
| Smoke-test base local completion (`abi complete`) | `complete-base` |
| Drive the SEA self-learning loop (`abi complete --learn`) | `sea-learn-loop` |
| Plan + train an agent (`abi agent plan` / `train`) | `agent-plan-train` |
| Deep-dive AI routing/constitution/SEA | `abi-superpower-ai`, `abi-superpower-sea`, `abi-superpower-constitution` |
| Run SEA learning, or inspect the REPL's session-only preference | `sea-learning-controller` (`abi complete --learn`; `/sea` is metadata only) |
## Facts that constrain any AI plan (from source, not marketing)
- Default model `claude-fable-5`; `crates/abi-ai/src/models.rs` is the single
source of truth for ids/aliases/provider routing (mod/stub parity).
- Router `selectBestProfile` ties resolve `abbey > aviva > abi`; neutral input
routes to `abi`. `analyzeSentiment` is **prefix-only** keyword matching.
- Constitution audit is **observability-only, not a gate** — sets
`audit_passed` / `audit_vetoed` / `escore`; safety+privacy hard-veto only
when either scores < 0.5. Matching is case-insensitive **substring** (infix),
so "harm" fires on "harmless" — it cannot detect novel harm patterns.
- EMA weights persist **only** on the `--learn`/SEA path; plain `complete`
re-runs sentiment each turn with no EMA persistence.
## Honest boundary
No energy-efficiency, accuracy, or QPS claims without a repo test/benchmark
proving them (see `docs/contracts/external-claims-audit.mdx`).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |