mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-11-06 05:17:15 +00:00
16 lines
1.8 KiB
YAML
16 lines
1.8 KiB
YAML
name: hermes-autonomous
|
||
repo: https://github.com/xiaok/hermes-autonomous
|
||
sha: 431c5fa31eceb5341504ef71db5b4bdb8d493036
|
||
description: "Proactive follow-ups: reads the recent conversation, finds things worth returning to (care / utility / risk / curiosity) and comes back at the right moment — a check on the weather before a trip, the price of something bought yesterday, or just \"did your throat get better?\". Event-driven rather than a daily digest: a cheap no-LLM gate runs on the cron tick and only wakes the agent when something is actually due. Sending goes through followups.py send, which enforces the 90-minute cooldown, the 23:00–08:00 quiet window, muted categories, at most three items and six lines per message, and records the receipt immediately. Disclosure — it reads the Hermes session database (~/.hermes/state.db) read-only: only sessions owned by the configured owner, DMs by default, with group/thread/forum/channel sessions (including other members' messages) only after setup --include-groups; it reads the cron job store (~/.hermes/cron/jobs.json) read-only when doctor or setup runs, and writes only its own plugin-data directory (mode 0600). The install step copies two helper scripts into ~/.hermes/scripts/; doctor --fix edits only this plugin's own cron job via hermes cron edit and never creates or deletes jobs. During a review the agent runs the plugin's helper through its terminal tool, and followups.py send delivers the DM with hermes send; the optional offline scripts/decide.py calls hermes -z. No credentials are read and there is no telemetry."
|
||
maintainer: xiaok
|
||
tier: community
|
||
category: automation
|
||
requires_hermes: ">=0.22.2"
|
||
version: "0.7.1"
|
||
platforms: []
|
||
capabilities:
|
||
provides_tools: []
|
||
provides_hooks: []
|
||
provides_middleware: []
|
||
requires_env: []
|