Petasos Fmarket NAV.

A Hermes Agent skill — a package that lets the open-source Hermes AI assistant do a job for you — that watches your mutual-fund investments. Every evening it reads the day’s published price from Fmarket, appends it to a small series file inside Obsidian, and speaks up only when your own rules say it’s time to act — so investment decisions come from fresh numbers, not from opening five apps.

Where it earns its keep

• “Should I add more today?” — when the fund’s price drops 10% or 17% below your average cost, it says so — once per level, never nagging.
• Every evening — the day’s price lands in your vault automatically; the series only ever grows, and nothing is overwritten.
• “Is it time for my monthly top-up?” — on the days you chose it reminds you before the order cut-off.
• “How is the fund doing?” — one generated page answers with fresh numbers: latest price, month and year change, and gain or loss against your cost.

How it’s built

• Your rules, your data. Which funds you track, what you paid, your limits — all in one plain file in your vault. Adding a fund is one entry, no code changes.
• Nothing happens by accident. A preview shows exactly what would change before anything is written; a malformed rule file stops everything loudly, naming the line.
• Quiet by design. If nothing needs your attention, it says nothing at all — news arrives only when your own thresholds are crossed.
• Private numbers. What you own and what you paid live only in your vault; the public code ships fake examples, and a privacy test fails the build if a real figure ever sneaks in.

Proven by tests

The whole suite runs offline — no real Fmarket calls — with 19 tests, including a privacy sweep over every file. Each live-API quirk found in the wild became a permanent test case, so the same mistake cannot return silently.

It needs only Python and one small library.