Barnaby Robson

My Public Agent Skills

Agentic engineering for knowledge workers.

I’ve created around 90 skills for my own work. These are eight I’ve chosen to publish to help other knowledge workers organise context and coordinate AI agents for analysis, research, writing and presentations. The collection includes my main AGENTS.md instruction file, which directs agents to the context and skills each task needs.

Complete collection

All eight skills, with the README and AGENTS.md.

Start with AGENTS.md

The main AGENTS.md is the entry point for the whole setup. It sets shared rules and directs the agent to the project or vault instruction file, the relevant skill and the session knowledge needed for the task.

Use the public example as a starting point and adapt its paths and rules to your machine. Keep local instructions at each project or vault root so the agent reads the rules for the place where it works. The collection includes this public instruction file.

Read the public AGENTS.md

AGENTS.md routes an agent to local instructions, skills and session knowledge
View full AGENTS.md illustration

Public skills

Source on GitHub

recall

Past agent sessions condensed into decisions, open work and a next action
View full recall illustration

Finds past work in session notes where available, with native transcripts for missing detail or a setup without notes

Graph mode uses networkx and pyvis. Session search uses your notes or native transcripts.

Download ZIP 28 KB

retrospective

Session evidence reviewed to identify repeated problems and propose small fixes
View full retrospective illustration

Reviews a session or a bounded recent sample, verifies repeated problems and proposes small fixes. Uses recall for historical evidence

Install recall for reviews of earlier sessions. Reviews of the current conversation run alone.

orchestrator

A coordinator routes work to agents and carries context through handoff briefs
View full orchestrator illustration · Model labels are illustrative. Your configured routing table selects the models.

Coordinates workers, routes each job to a model and effort level, and hands a session on before its context runs out. Owns the routing table, the category rule and the resolver

Install the cmux companion skill and PyYAML. Configure the routing table before spawning workers.

Download ZIP 501 KB

cmux

Scattered terminal windows organised into one cmux workspace with a saved session note
View full cmux illustration

A skill I created for CMUX, my command-line interface of choice. It starts and closes workspaces, writes session notes and uses the orchestrator skill to route work.

Install the cmux application and the orchestrator companion skill.

Download ZIP 20 KB

skill-manager

Long instructions reorganised into a focused SKILL.md with references, scripts, assets and validation checks
View full skill-manager illustration

Writes and reviews new skills: description rules, a pre-ship checklist, a validator and a whole-tree audit

The validator needs Bash. The whole-tree audit needs the Claude Code Workflow tool.

Download ZIP 18 KB

Install a skill

A skill is a folder with a SKILL.md in it. Put the folder where your coding agent looks for skills.

  1. Extract the ZIP. An individual archive contains one skill folder. The complete collection contains the skill folders under public-skills/skills/.
  2. Put each skill folder in ~/.claude/skills/ for Claude Code or ~/.codex/skills/ for Codex. Preserve existing folders and review their contents before replacing them.
  3. Install the companion skills and prerequisites listed above. For PyYAML, run python3 -m pip install --user PyYAML. Configure the routing table in SKILL.md and references/routing-metadata.yaml before spawning workers.
  4. Restart your agent so it reads the skill.

Read the full setup guide. The collection ZIP includes the same README.