Task Router Skill
Task Router
Advanced LYGO-MINT Operator Suite (v2): canonicalize *multi-file* packs, generate per-file + bundle hashes, write append-only + canonical ledgers, produce machine-readable multi-platform Anchor Snippets, and verify third-party packs. Built for LYGO operators who want dependable, receipts-first truth anchoring across MoltX/Moltbook/X/Discord/4claw.
请根据来源仓库和你的 AI 工具说明安装此技能。先阅读技能文件和依赖要求,再添加到工作流。
SkillHub CLI
skillhub install deepseekoracle/lygo-mint-operator-suite打开来源仓库,查看说明和兼容环境后,按来源提供的命令或步骤安装。
将来源链接提供给你的 Agent,让它先读取说明、核对依赖,并向你说明将执行的安装步骤。
This skill turns an aligned Champion pack (or any prompt/workflow pack) into a verifiable artifact:
Read the full process doc: references/process.md.
All scripts are local and never print secrets.
python scripts/mint_pack_v2.py --input reference/CHAMPION_PACK_LYRA_V1.md --title "LYRA Pack" --version 2026-02-09.v1
python scripts/mint_pack_v2.py --input skills/public/lygo-champion-kairos-herald-of-time --title "KAIROS Pack" --version 2026-02-09.v1
python scripts/verify_pack_v2.py --input ./some_pack_folder --pack-sha256 <hash>
python scripts/bundle_pack_v2.py --input ./some_pack_folder --out tmp/pack.bundle.zip
python scripts/make_anchor_snippet_v2.py --pack-sha256 <hash> --title "..." --platform moltx
python scripts/backfill_anchors.py --hash <64-hex> --channel moltbook --id <post-id-or-url>
state/lygo_mint_ledger.jsonl
state/lygo_mint_ledger_canonical.json
reference/CHAMPION_PROMPT_CORE_TEMPLATE_V1.md
reference/CHAMPION_PACK_PUBLISH_CHECKLIST.md
这些信息来自收录时的来源分析,可能滞后;使用前请核对最新来源文件。
# LYGO-MINT OPERATOR SUITE (v2) This skill turns an aligned Champion pack (or any prompt/workflow pack) into a **verifiable artifact**: - canonical form - deterministic hash - ledger receipts - portable Anchor Snippet (paste anywhere) ## Workflow (high-level) 1) Create/align the pack (e.g. Champion alignment system). 2) Run verifier: - canonicalize - hash (SHA-256) - write ledgers - emit Anchor Snippet 3) Post Anchor Snippet anywhere. 4) Backfill anchor IDs into the ledger. Read the full process doc: `references/process.md`. ## Commands (scripts) All scripts are local and never print secrets. ### 1) Mint a pack (file OR folder) → manifest + hashes + ledgers + snippets - `python scripts/mint_pack_v2.py --input reference/CHAMPION_PACK_LYRA_V1.md --title "LYRA Pack" --version 2026-02-09.v1` - `python scripts/mint_pack_v2.py --input skills/public/lygo-champion-kairos-herald-of-time --title "KAIROS Pack" --version 2026-02-09.v1` ### 2) Verify a pack against an anchor snippet or a known hash - `python scripts/verify_pack_v2.py --input ./some_pack_folder --pack-sha256 <hash>` ### 3) Create deterministic bundle (zip) for distribution - `python scripts/bundle_pack_v2.py --input ./some_pack_folder --out tmp/pack.bundle.zip` ### 4) Generate multi-platform anchor snippets - `python scripts/make_anchor_snippet_v2.py --pack-sha256 <hash> --title "..." --platform moltx` ### 5) Backfill anchors (post IDs/links) - `python scripts/backfill_anchors.py --hash <64-hex> --channel moltbook --id <post-id-or-url>` ## Ledgers (workspace state) - Append-only: `state/lygo_mint_ledger.jsonl` - Canonical (dedup): `state/lygo_mint_ledger_canonical.json` ## References - Core template: `reference/CHAMPION_PROMPT_CORE_TEMPLATE_V1.md` - Publish checklist: `reference/CHAMPION_PACK_PUBLISH_CHECKLIST.md`
Task Router
Register, communicate, and earn on the x402hub AI agent marketplace. Use when an agent needs to register on x402hub, browse or claim bounties, submit deliverables, send messages to other agents via x402 Relay,...
Real-time event bus for AI agents. Publish, subscribe, and share live signals across a network of agents with Unix-style simplicity.
OpenClaw integration guidance for CAPTCHAS Agent API, including OpenResponses tool schemas and plugin tool registration.