AGENT SKILL

learning-loop

Explain and control the stellar-build learning loop - the system that improves your skills from how you use them. Use when the user says "learning loop", "how does the learning loop work", "what is the learning loop", "self-improving skills", "stellar-loop", "show my skill usage", "is tracing on", or wants an overview / status of the local trace-capture + optimize + bench cycle.

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 6
Forks
3

01 前言 Front Matter

name
learning-loop
description
Explain and control the stellar-build learning loop - the system that improves your skills from how you use them. Use when the user says "learning loop", "how does the learning loop work", "what is the learning loop", "self-improving skills", "stellar-loop", "show my skill usage", "is tracing on", or wants an overview / status of the local trace-capture + optimize + bench cycle.

03 SKILL.md 正文

What the learning loop is

stellar-build ships with a local learning loop: the more you use your skills, the better they get - without anything leaving your machine. It mirrors OpenJarvis's jarvis optimize skills / jarvis bench skills cycle, adapted for markdown skills.

Three moving parts:

  1. Capture - a skill use is a multi-turn span, so capture is decoupled from "skill became active". A PostToolUse:Skill hook marks which skill is resident; a Stop hook records each turn's real assistant output; and a SessionEnd hook closes the span with a summary, counting it as one activation. Everything lands in ~/.stellar/traces/<day>.jsonl. Fully local. Opt out with STELLAR_BUILD_NO_TRACE=1.
  2. Optimize (/optimize-skills) - compiles those traces into sharper skills: tightened triggers/instructions and few-shot examples mined from your own successful runs. Every edit is backed up and reversible.
  3. Bench (/bench-skills) - scores skills on held-out prompts so you can measure whether an optimization actually helped.
   use skills  ──▶  ~/.stellar/traces/*.jsonl
                          │
              /optimize-skills  (DSPy-style compile)
                          │
                          ▼
            sharper ~/.claude/skills/<skill>/SKILL.md   ◀── stellar-loop restore (undo)
                          │
                /bench-skills  (measure the lift)

What to do when invoked

  • Overview / "how does it work" → explain the three parts above, then point to the two action skills and the CLI.
  • "status" / "is tracing on" / "show my usage" → run the CLI and relay the output:
    • stellar-loop status - install + capture + hook state, trace/optimization counts
    • stellar-loop stats - per-skill usage from local traces
    • stellar-loop tail - most recent trace records The CLI is installed at ~/.stellar/bin/stellar-loop (add ~/.stellar/bin to PATH, or call it by full path).
  • "optimize" / "improve my skills" → hand off to the optimize-skills skill.
  • "benchmark" / "did it help" → hand off to the bench-skills skill.
  • "undo" / "restore" → stellar-loop restore <skill> reverts the last optimization for that skill from its backup.

Data & privacy

  • Everything lives under ~/.stellar/ (traces/, backups/, bench/, optimizations.jsonl). Override the root with STELLAR_HOME.
  • Nothing is transmitted anywhere - no API calls, no telemetry. The "optimizer" is your own agent session reasoning over local files.
  • Result previews are truncated and examples are generalized before they ever land in a skill, so traces and learned blocks should not accumulate secrets. Wipe anytime with stellar-loop clear (--all also clears backups).

Quick reference

Goal Do this
See if it's working stellar-loop status
See what you use most stellar-loop stats
Make skills sharper /optimize-skills (or stellar-loop optimize)
Measure the lift /bench-skills (or stellar-loop bench)
Undo an optimization stellar-loop restore <skill>
Turn capture off export STELLAR_BUILD_NO_TRACE=1
Delete all traces stellar-loop clear

04 同仓库更多 Skills

WritzProtocol
WritzProtocol/writz

zk-proofs

Zero-knowledge proofs and privacy patterns on Stellar. Covers Groth16 verification in smart contracts via BLS12-381 host functions (CAP-0059, available), the BN254 + Poseidon proposals (CAP-0074/0075, status-sensitive), and concrete toolchain walkthroughs for Circom (on-chain verifiable today), Noir, and RISC Zero (attestation pattern until BN254 lands). Use when building privacy-preserving applications, ZK-verifier contracts, or wiring a proving toolchain to Stellar.

★ 6
WritzProtocol
WritzProtocol/writz

tyler-architect

System architect and technical design leader. Use when the user asks to talk to Tyler or requests the architect.

★ 6
WritzProtocol
WritzProtocol/writz

stellar-help

Analyzes current state and user query to answer BMad questions or recommend the next skill(s) to use. Use when user asks for help, bmad help, what to do next, or what to start with in BMad.

★ 6
WritzProtocol
WritzProtocol/writz

stellar-competitive-landscape

Map the competitive landscape for a Stellar project idea. Use when a user says "who are my competitors on Stellar", "competitive analysis Stellar", "what already exists in this space on Stellar", "show me similar Stellar projects", "is anyone else building this on Stellar", or "map the Stellar landscape for X". Queries the 728-project LumenLoop ecosystem database and Electric Capital developer activity to rank competitors by SCF funding history and repo activity.

★ 6
WritzProtocol
WritzProtocol/writz

standards

Stellar standards, ecosystem, and reference. Covers SEPs (Stellar Ecosystem Proposals), CAPs (Core Advancement Proposals), and a quick map for picking the right standard for wallets, anchors, payments, deposits/withdrawals, federation, deep links, and KYC. Also bundles ecosystem references (DeFi protocols, dev tools, wallets, infra, community projects), curated documentation links, and MCP servers (live tools such as Raven). Use when you need to know which SEP applies, or want a starting point for ecosystem integrations, official docs, or live MCP tooling.

★ 6
WritzProtocol
WritzProtocol/writz

smart-contracts

Stellar smart contract development (Rust, soroban-sdk). Entry point with project setup, contract anatomy, and build/deploy workflow, routing to three companion files in this directory - development.md (storage/TTL, authorization, cross-contract calls, tokens, events, errors, upgrades, fees, troubleshooting), testing.md (unit, fuzz, property, fork, mutation, integration), and security.md (vulnerability classes, checklists, tooling, audits). Use when writing, testing, reviewing, securing, debugging, or shipping Stellar smart contracts, including anything the user calls "Soroban" - Soroban contracts, soroban-sdk, Soroban auth/storage/TTL errors, SEP-41 tokens, or SAC integration from contract code.

★ 6