Ipfs Pin Stage

通过本地 kubo 节点添加并固定文件，维护带 TTL 的账本；支持刷新、取消固定、按 TTL 衰减并执行仓库回收，附带每日 systemd 定时清理和状态查询。

paradoxfuzzle

热度 17.01

最近更新 2026-10-09 21:29:04

分类：自动化依赖：代码执行 / 读写文件 / 第三方 API应用场景：工作流 / 自动化脚本 / DevOps

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add ipfs-pin

Metadata

name: ipfs-pin
description: IPFS content persistence for OpenClaw via a local kubo node. Adds files to IPFS, pins them so they stay robust while active, and tracks a per-file TTL in a ledger. `decay` unpins entries not referenced within their TTL and garbage-collects the repo; a systemd user timer runs it daily so unused content decays naturally without manual cleanup. Reports only CIDs and status the ipfs CLI actually returned.
user-invocable: true
metadata: {"openclaw":{"emoji":"🌐","requires":{"bins":["ipfs"],"files":[],"config":[]}}}

ipfs-pin skill

Content is added to the local kubo node (API on 127.0.0.1:5001) and pinned. Content-addressing means every file gets a stable CID that works across the IPFS network; pinned entries are also served from local disk so they survive offline. The ledger at ~/.openclaw/ipfs-pin/ledger.json tracks each CID with a TTL; the decay sweep (run daily by an ipfs-pin-decay.timer systemd user unit) unpins entries past TTL and runs ipfs repo gc, so unused content decays rather than accumulating forever.

Only cite CIDs or status the ipfs CLI returned — never invent one. If the daemon is down, say so (status reports it); don't guess.

Setup (done once)

kubo binary: ~/bin/ipfs (kubo 0.43.1, verified via upstream .sha512)
daemon: systemctl --user status ipfs (unit in ~/.config/systemd/user/ipfs.service)
decay sweep: systemctl --user status ipfs-pin-decay.timer

Command reference

Run via bash: ~/.openclaw/plugin-skills/ipfs-pin/ipfs_pin.py <cmd> .... Add --json right after the program name for structured output.

Command	Purpose
`add <path> [--name N] [--ttl H]`	Add+pin a file; TTL refreshes on re-add. Default TTL 72h (`config --default-ttl-hours` to change).
`ls`	Show tracked CIDs with hours remaining
`touch <cid>`	Refresh `last_referenced` on a CID you're still using
`rm <cid> [--no-gc]`	Unpin immediately and drop from ledger
`decay [--dry-run] [--quiet]`	Unpin everything regretted past TTL + `repo gc`
`get <cid> [-o PATH]`	Retrieve by CID from local node (cat to stdout or write file)
`status`	Daemon up? repo size, pin count, tracked vs untracked
`config [--default-ttl-hours H]`	Show or set default TTL

Decay semantics

A file added today with default TTL decays ~3 days from last reference.
Re-adding the same content (identical bytes = same CID) refreshes its TTL instead of duplicating.
Explicitly touch a CID during work to keep it alive without re-adding.
decay --dry-run previews what would be unpinned without touching anything.
Manual ipfs add (outside this skill) still pins on the node — status flags these as "untracked" so they don't silently miss the sweep.

One-command decay check

~/.openclaw/plugin-skills/ipfs-pin/ipfs_pin.py decay --dry-run