Independent methods. Inspectable results. Agent skills, local tools and weekly intelligence.

Agent Optimization Memo · $29 / month

Make sense of the week. Put the useful parts to work.

For people building with AI agents who want more than a pile of links. Get a curated weekly packet of agent findings, studies, repos and skills, with practical reviews, clear recommendations and original skills included.

Subscribe · $29/monthRead the free packet preview

A useful shortlist, not another backlog

  • What changed: selected agent releases, studies, repos and skills, dated and linked to their primary sources.
  • What to do: practical reviews, who each finding fits, and a clear next step.
  • What to use: original skills with supporting files, included in the packet.
  • What to test next: bounded experiments and changes worth watching.

Inside the November 12 packet

13 reviewed developments · 2 original skills · 3 practical experiments.

Seven findings from November 6–12, plus six clearly labeled recent carryovers. Includes four studies, two engineering articles and seven framework, SDK or skill releases.

  • Hermes and OpenClaw: plugin freshness, skill migrations, private-result isolation and failure reporting.
  • Claude Code, OpenAI Agents SDK and LangGraph: latency work, alpha migration choices and resume boundaries.
  • Research into practice: what recent studies suggest testing in your own agent setup.

Two original skills, included

Agent Run Audit: inspect locally supplied run metadata for large outputs, repetition, slow steps and token hotspots.

Skill Trigger Lab: score observed skill activation separately from task success. See when the right skill was missed or an irrelevant skill fired.

The included ZIP also retains the original paid launch essay in a clearly labeled previous-edition folder.

Portable SKILL.md instructions, local Python helpers and worked examples. Bring your run metadata and skill observations; get structured findings you can act on.

Free preview · November 12 packet

A missing plugin may be a catalog problem.

Hermes Agent v0.23.1 · Check → test patch

The maintainers report that a catalog-size limit caused plugin installs to fall back to an older bundled list. Version 0.23.1 raises that limit. A missing plugin or outdated pin may therefore need a catalog/version check, not a rewritten prompt.

Our recommendation: confirm your installed version and the intended upstream plugin version first. Test the patch in a reversible profile, then run one harmless workflow that uses the plugin. Keep the old environment until the result is right.

Who it fits: Hermes users troubleshooting plugin discovery or stale versions. The release notes say standalone desktop installers are unchanged in this patch.

Source-based review · November 12. Read the maintainer release notes ↗

See what an included skill produces

Get your skills to the right task.

Skill Trigger Lab turns your activation observations into a focused review: which requests activated the intended skill, which missed it, and which triggered an irrelevant skill. The report names the case IDs to revisit and keeps activation scoring separate from task success.

Use the results to refine your skill’s description and examples, then compare the revision on fresh requests. The included local scorer works with observations you provide.

The full packet adds 12 more reviewed findings, both original skills and three practical experiments.

Your weekly packet

A new issue appears in your subscriber archive every Thursday at 17:00 UTC. Read it there or have your agent retrieve it. The subscription includes the archive and new issues through your paid period. Keep editions you download.

Why pay when the sources are public?

You can follow the original links yourself. The paid work is selecting what matters, reviewing the tradeoffs, turning findings into concrete next steps, and packaging original skills you can use.

How are recommendations developed?

Each review connects primary sources to practical choices: who a change fits, what to try and what to watch. Source links and review scope are included so you can follow the reasoning.

Does it include the other shop packs?

Shop packs are separate one-time purchases. The original skills named in each Memo packet are included with the subscription.

Use the Memo in your own workflow

Commercial use of downloaded current editions and original skills in your own or your organization’s work, including AI-agent and client projects. Share your own outputs; do not redistribute or resell the current pack. Downloaded editions remain usable after cancellation. Prior grants and separately identified historical material keep their existing rights.

Read the current skills license

Access and renewal

Save your private access key during checkout. Use the same browser or restore access with that key on another device or agent. Keep a copy somewhere safe; AOM cannot automatically recover a lost key. Manage payment and cancel renewal through Link.

Applicable tax and any local-currency total are shown before payment. Refund requests follow Link’s refund policy and purchase terms. New downloads require a current eligible paid subscription.