witcheer's Wingtips for Hermes Agent

Wingtips

1. #105 · trust: untrusted

   Oct 18, 2026 · Security & approvals · CLI / Desktop / Gateway

   want your Hermes Agent to ask before it changes anything through an MCP server?

   trust: untrusted

   add it under that server in config.yaml. tools the server marks read-only run as normal, and every other tool waits for your /approve.

   docs: https://hermes-agent.nousresearch.com/docs/reference/mcp-config-reference#server-keys

   SourcesHermes docs: Mcp Config Reference

2. #83 · approvals.denial_breaker_threshold, the smart-approval denial breaker

   Sep 25, 2026 · Security & approvals · CLI / Desktop / Gateway

   when the smart-approval reviewer denies a command, Hermes Agent can try a variation of it.

   this key counts those denials in a row, per session. at 3, the deny message tells the agent to stop, report what was blocked, and ask you to run it yourself or /approve it.

   lower it to stop sooner, or set 0 to turn it off.

   docs: https://hermes-agent.nousresearch.com/docs/user-guide/configuration#denial-circuit-breaker

   SourcesHermes docs: Configuration

3. #81 · unauthorized_dm_behavior, what your bot says to someone you have not approved

   Sep 23, 2026 · Security & approvals · Gateway

   when someone you have not approved messages your Hermes bot, it answers with a one-time pairing code, so you can let them in from the CLI.

   new in Hermes Agent: if the bot is only for you, it can say no instead.

   unauthorized_dm_behavior: decline

   one short, polite decline, then silence toward that sender for a day.

   docs: https://hermes-agent.nousresearch.com/docs/user-guide/configuration#unauthorized-dm-behavior

   SourcesHermes docs: Configuration

4. #78 · approvals.smart_policy, your own rules for the command reviewer

   Sep 20, 2026 · Security & approvals · CLI / Desktop / Gateway

   when Hermes Agent wants to run a terminal command that looks risky, a second model reviews it first. it reads the command and answers with one word: approve, deny, or escalate, which means ask you.

   this key lets you add your own rules to that reviewer's instructions.

   "always escalate anything that modifies /etc." "approve docker compose restarts in ~/deploys, they are routine here."

   your rules are read as instructions from you, never as part of the command it is judging.

   Copy code

   approvals:
     smart_policy: |
       Always ESCALATE commands that modify anything under /etc.
   

   empty by default. the reviewer itself is on by default (approvals.mode: smart).

   docs: https://hermes-agent.nousresearch.com/docs/user-guide/configuration#smart-approvals

   SourcesHermes docs: Configuration

5. #69 · tool_loop_guardrails

   Sep 11, 2026 · Security & approvals · CLI / Desktop / Gateway

   Hermes Wingtips #69: tool_loop.guardrails

   Show full tip

   your agent opens a file, reads it, decides to start over, and reads the same file again.

   Hermes Agent keeps count of that inside one turn. it watches for three patterns: the same call failing twice in a row, the same tool failing on new arguments, and a call that comes back with the same result while nothing has changed. when a count is reached it writes a warning into the tool result itself, so the model reads it on the next step and changes course.

   in the CLI, the TUI and the Desktop app that is where it ends, because you are there to step in. gateway and cron runs go one step further and stop the turn, since nobody is watching. a stop ends the turn, not the session, and the agent tells you which pattern fired.

   hermes config set tool_loop_guardrails.hard_stop_enabled true

   docs: https://hermes-agent.nousresearch.com/docs/user-guide/configuration#tool-loop-guardrails

   SourcesHermes docs: Configuration

6. #41 · approvals suggest

   Aug 9, 2026 · Security & approvals · CLI

   Hermes Wingtips #41: approvals suggest

   approving the same command for the third time this week? your Hermes Agent already keeps that history.

   hermes approvals suggest mines your past approvals and turns the recurring ones into an allowlist proposal, so the prompts you always say yes to stop coming back.

   it proposes, you pick, nothing is written on its own.

   below, a real run on my own agent this morning.

   SourcesHermes docs: Security

7. #40 · give your local agent two commands, not root

   Aug 7, 2026 · Security & approvals · CLI / Gateway

   Hermes Wingtips #40: give your local agent two commands, not root

   when I'm running benchmarks, my local Hermes Agent stops and starts the model server on its own, from cron, with no password. it still cannot touch anything else as root.

   the trick is a sudoers rule scoped to exactly one service: NOPASSWD applies to systemctl start and stop for that unit, nothing more.

   still test the fence from the outside: sudo -n on any other service should still ask for a password.

   SourcesHermes docs: Security

8. #18 · keep auto-learning, review the writes

   Jul 10, 2026 · Security & approvals · CLI / Desktop / Gateway

   Hermes Wingtips #18: keep auto-learning, review the writes

   if you want Hermes to keep learning without changing memory or skills unchecked, turn on both approval gates:

   ~ /memory approval on ~ /skills approval on

   new writes wait for your yes instead of landing immediately.

   SourcesHermes docs: Memory