Skip to content
aiagentsetups.com

Setups

Shubham Saboo's agent squad

Shubham Saboo (@Saboo_Shubham_) · Catalog date October 16, 2026

Shubham Saboo describes a squad that started on OpenClaw and is moving to Hermes, both driven through Telegram. Monica grades the squad. A later post names a Hermes chief of staff and specialist agents who own domains. The agents help run the Awesome LLM Apps repository. The author still makes the final decision.

Evidence. Sourced, incomplete

A public source is linked. Some results, prompts, or operating details are still unknown. This catalog has not re-run the crew.

Goal
Keep a public repository moving, with the author making the final call
Use
Software
Platforms
OpenClaw, Hermes
Crew size
Multiple agents, count not stated
Difficulty
Not stated

Roles

The June 28, 2026 post names Monica and says each agent reviews its own work. The August 30, 2026 post names a Hermes chief of staff and specialist agents who own domains. It does not name those specialists or give a count. A 19-minute video is attached to the June post. This entry uses the post text, not a transcript of the video.

Constraints shared by every role

To be added.

  • Monica

    Owns. Reviews the whole squad every two weeks and grades them.

    Uses the shared constraints above.

  • Hermes chief of staff

    Owns. Named in the execution pillar of the August post, with specialist agents and CLIs.

    Uses the shared constraints above.

  • Domain specialists

    Owns. The August post says specialist agents own domains. Individual names and charters are not in the post text.

    Uses the shared constraints above.

Routines

  • Own-work review

    Monthly

    Each agent reviews its own work.

  • Monica's review

    Every two weeks

    Monica reviews the squad and grades them.

  • Daily signals

    Daily

    The agents send proactive daily signals so the author can stay up to date.

  • Cron and heartbeats

    Schedules and heartbeats; clock times are not given

    Agents run on cron and heartbeats instead of waiting for a prompt.

Guardrails

  • The agents do the work, and the author still makes the final decision.
  • The author sets goals, approves, and corrects when needed.

Handoffs

Work escalates to the author for the final decision. Monica reviews the other agents on a two-week cycle. The posts do not describe how a specialist hands a specific task to another specialist.

Recorded handoff
  • Monica
  • Hermes chief of staff
  • Domain specialists

Human operator

Work escalates to the author for the final decision. Monica reviews the other agents on a two-week cycle. The posts do not describe how a specialist hands a specific task to another specialist.

Results

What worked
The June post says these agents manage the open-source Awesome LLM Apps repository, which the post states has 115k+ GitHub stars, and that they send proactive daily signals.
What broke
To be added.
Lessons
To be added.

Prompts

To be added.

Source

Notes

The catalog date is the day this entry was drafted. The two source posts are dated June 28, 2026 and August 30, 2026.

The June post says both OpenClaw and Hermes are used through Telegram. The August post adds four pillars in the author's words: direction (the author sets goals, approves, and corrects), execution (Hermes chief of staff, specialist agents, and CLIs), learning (skills preserve what works, and memory stores taste), and feedback (outcomes and metrics improve the next run). The posts do not include a prompt, a full roster, what broke, or lessons. Difficulty is not stated.