New community skill: audit and critique your Agent OS profiles from inside Claude

[psenger]

New community skill: audit and critique your Agent OS profiles from inside Claude

I have been using Agent OS for a while now and kept running into the same friction: writing profiles feels productive, but there is no easy way to know whether the standards you are writing are actually doing useful work in the context window.

So I built a skill that audits them.

---

agent-os-profile-critique is a new community skill that loads structured audit checklists and quality criteria directly into Claude's context. Point it at a profile directory, a project install, or an enterprise profiles repo, and it produces a prioritised findings report: severity-tagged, anchored to file paths and line numbers, with concrete fixes rather than general advice.

A few things it catches that are hard to spot manually:

Standards that earn their place vs. ones that waste tokens. A standard that restates something Claude already handles well, with no code example and no opinionated rule, is deadweight in every context window. The skill identifies those and tells you why.

v2 artifacts in v3 installs. Subagents, legacy profile-config.yml locations, agents/ folders inside profiles. Flagged on sight, with migration steps.

Inheritance chain coherence. When you have a chain, the skill builds a contribution map showing which profile contributes each standard and where it is overridden. It surfaces generality leaks, override saturation, and cross-level conflicts that the file-by-file mechanism hides.

Every finding is tagged with its source: [ref] (loaded from skill references), [corpus] (from the model's training), or [both], so you know how much weight to give each one.

---

Install:

npx skills add psenger/ai-agent-skills --skill agent-os-profile-critique

Source and docs at https://github.com/psenger/ai-agent-skills (MIT).

---

Feedback, bug reports, and pull requests are welcome. If you have profile patterns that should be in the audit checklists, raise an issue at https://github.com/psenger/ai-agent-skills/issues and I will work them in.