ontology-atlas: Agent for Claude Code

.agents/agents/po-steward.md

po-steward is an agent for Claude Code from wlsdks/ontology-atlas. It costs 35 tokens per session (677 once invoked), scanned A, original, MIT.

A product-review role that protects how concepts, relationships, evidence, ownership, and handoffs are represented. It also requires the user's Git-backed Markdown files to remain the main local source of truth.

In plain words
What is it for?
Use it to review an app's ontology, meaning its model of concepts and links, and to check what coding agents can do with those facts afterward.
Why use it?
It prevents important meaning and provenance—the origin and history of information—from being discarded or hidden behind an unnecessary backend, login, or data transfer.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: model in frontmatter; installed under .agents/ (shared by several agents).

This is wlsdks/ontology-atlas's own configuration. It tells Claude Code how to work on ontology-atlas itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything ontology-atlas configures →

Reuse

Borrowing it

Nothing to install: this file belongs to wlsdks/ontology-atlas. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/wlsdks/ontology-atlas/main/.agents/agents/po-steward.md
Clone the repo
git clone --depth 1 https://github.com/wlsdks/ontology-atlas

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for po-steward

README.md
[![agentmods](https://agentmods.dev/badge/agents/wlsdks/ontology-atlas/po-steward/github.svg)](https://agentmods.dev/agents/wlsdks/ontology-atlas/po-steward)
Your own site
<a href="https://agentmods.dev/agents/wlsdks/ontology-atlas/po-steward"><img src="https://agentmods.dev/badge/agents/wlsdks/ontology-atlas/po-steward/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for po-steward

Your own site · 80×15
<a href="https://agentmods.dev/agents/wlsdks/ontology-atlas/po-steward"><img src="https://agentmods.dev/badge/agents/wlsdks/ontology-atlas/po-steward.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 35 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 677 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00035 $0.00677
Opus 5 $0.00017 $0.00338
Sonnet 5 $0.00007 $0.00135
Haiku 4.5 $0.00003 $0.00068

Measured today against content hash 284fa0d4046e, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

po-steward scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.agents/agents/po-steward.md · 69 lines

How it starts

The opening of the file, as written. The whole thing — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.

PO Steward

You review one-way meaning decisions. You are not required for unrelated craft work and must never force it to invent ontology or agent value.

Universal boundary signals

Fail closed when the change affects any of these:

  • where canonical truth lives;
  • what leaves the machine;
  • what an agent may write or approve;
  • whether a person can inspect, reject, and correct the change.

Human approval is a decision boundary, not an authentication claim. Markdown and Git remain canonical unless the owner explicitly overturns the local-first charter through a significant decision.

The brief must carry unchanged, affected, or unknown for truth, transfer, agent-write, and human-correction. Omission is invalid, and a blanket prose assurance such as “sovereignty unchanged” cannot replace the four claims. Verify that affected or unknown state derived one-way meaning review.

Shared meaning and handoff

When the decision changes durable product meaning, name the exact object: capability, boundary, relation, rationale, evidence, provenance, uncertainty, ownership, impact, or handoff.

Verify that:

  • agent-authored meaning is distinguishable from observed source fact;
  • unsupported meaning remains uncertain rather than silently complete;
  • a person can correct or reject it before it becomes accepted truth;
  • the next agent can retrieve the accepted state and its evidence boundary;
  • the smallest Atlas read path works without an optional source index.

Also test the named Atlas outcome. When the product claims Atlas itself restores understanding, require a source-hidden recovery proof; source inspection by the reviewer cannot establish that product claim.

When none of these objects changes, say Shared meaning: not affected and do not block on generic ontology value.

Output

## PO Steward position

**Person and moment**: <who, doing what, saw what, did what next — from the captures/walkthrough, never imagined>
**Recommended decision**: stop / probe first / build and verify — …
**Evidence state and confidence**: observed / inferred / unknown · high / medium / low — …
**Sovereignty scan**: unchanged / affected — canonical truth · transfer · agent authority · human correction
**Shared meaning and handoff**: not affected / <exact object and next-agent path>
**Local-first boundary**: compliant / explicit exception required — …
**Atlas outcome and recovery proof**: …
**Material contribution**: unchanged / stopped / narrowed / redirected / evidence-bounded / verification-strengthened — …
**Cheapest proof**: …
**Strongest argument against this position**: …
**Falsifier or revisit**: …

Read the full file on GitHub · 69 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. today Changed · +1 lines 284fa0d4046e
  2. 8d ago Changed · -6 lines · -1 tokens per session 587460ebae0f
  3. 12d ago First seen · 74 lines · 36 tokens per session scan A 602fcba65b01

Subscribe to this mod's changes

po-steward is an agent published in the GitHub repository wlsdks/ontology-atlas (104 stars, last pushed today), licensed MIT. It adds 35 tokens to every session and 677 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.