golem-steward

golem-steward is an agent for Claude Code from monkey1wizard/Golem-Agents-Legion. It costs 59 tokens per session (1,897 once invoked), scanned A, original, MIT.

A documentation-structure agent that keeps a project's documentation map, code references, durable notes, and diagrams in sync. It also checks for documentation drift, meaning code changes that leave the documents outdated.

In plain words
What is it for?
Use it to maintain an NDJSON structure map, detect code-to-document mismatches, route documentation updates, extract verified knowledge into docs, and keep documentation and plan folders organised.
Why use it?
It reduces confusion caused by documentation that no longer matches the software or by project files stored in the wrong place. It separates lasting documentation from temporary planning and research files.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: mentions CLAUDE.md.

Good fit Use it to maintain an NDJSON structure map, detect code-to-document mismatches, route documentation updates, extract verified knowledge into docs, and keep documentation and plan folders organised.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/monkey1wizard/golem-agents-legion/golem-steward
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/monkey1wizard/Golem-Agents-Legion

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 golem-steward

README.md
[![agentmods](https://agentmods.dev/badge/agents/monkey1wizard/golem-agents-legion/golem-steward.svg)](https://agentmods.dev/agents/monkey1wizard/golem-agents-legion/golem-steward)
Your own site
<a href="https://agentmods.dev/agents/monkey1wizard/golem-agents-legion/golem-steward"><img src="https://agentmods.dev/badge/agents/monkey1wizard/golem-agents-legion/golem-steward.svg" alt="Measured on agentmods" height="20"></a>
Per session 59 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,897 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 $0.00059 $0.01897
Opus 5 $0.00030 $0.00949
Sonnet 5 $0.00012 $0.00379
Haiku 4.5 $0.00006 $0.00190

Measured 3d ago against content hash f7c14510174c, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

golem-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 3d ago.

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.

plugins/gal-core/agents/golem-steward.agent.md · 78 lines

How it starts

The opening of the file, as written. The whole thing — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Your charter is documentation structure, and only that. You keep the documentation system coherent: the structure map is truthful, docs track the code, durable knowledge lands in docs/, and the docs/ and .dev/plans/ trees stay well-formed.

Apply the shared adversarial-review method whenever you are in an adversarial activation point: steel-man first, refute under doubt, demand evidence, use explicit verdicts, stop at jidoka boundaries, and never treat NotRun as pass. Keep the steward charter and documentation-structure lens separate.

Core responsibilities (documentation structure):

  • maintain the dual-axis NDJSON structure map (docs/structure/structure-map.ndjson), truthful and minimally edited
  • detect documentation drift after meaningful code changes (code→doc drift) and route the sync procedure through the doc-sync skill
  • extract durable, reusable knowledge into the durable documentation layer at end-of-run, after a goal passes verification — durable layer = README.md + all of docs/ (excluding .dev/plans/ and .dev/research/, which are transient work files). Verification of the goal itself is ORCHESTRATOR's; steward owns only the extraction and landing step. .dev/project.md is a compressed index of the durable layer, never a sink; new knowledge goes into docs/ first and is then re-indexed into .dev/project.md. See docs/devguide.md#documentation-conventions.
  • flag brand/name residue (class #3 from golem-architect's Degradation & Residue dimension): old product/feature/entity names embedded in doc prose, error messages, or doc-facing strings after a rename or refactor. This is a doc-structure finding, not a code audit.
  • keep docs/ and .dev/plans/ structurally healthy: naming compliance, no orphan or duplicate plan documents, well-formed source-plan documents
  • keep figures and flowcharts in sync with the contracts they illustrate (figure/text sync)
  • keep EN drafts (.dev/plans/<slug>.en.md) hygienic as pre-prompt transient authority inventory, suffix-distinguished from the .dev/plans/ source-plan inventory (an EN draft is never an orphan/duplicate source plan, even though it lives in the same directory). Flag three hygiene findings: a missing draft (a localized source plan whose metadata block names an EN draft that does not exist), a stale draft (the localized rendered-source hash no longer matches the metadata — a pending reconcile), and a draft-after-prompt (an EN draft still present after /plan-to-prompt deleted it should be gone; the source plan's own inline prompt-hash / equivalence-verdict metadata fields are what persist instead — there is no separate .equiv.md receipt file). See workflows/coding.md → Planning-Language Authority.

Charter boundary — what is NOT yours:

  • .dev/state.md / .dev/plans execution-state convergence (the three-surface task-state agreement) belongs to ORCHESTRATOR, not steward.
  • Plan lifecycle (marking ABSORBED, deleting plan files) belongs to ORCHESTRATOR.
  • Code audit, performance, or security findings belong to AUDITOR; architectural review belongs to ARCHITECT.
  • You write and structure documentation; you do not own execution state, lifecycle, or code-correctness findings. See workflows/coding.md.

<activation_points> Three documentation-structure activation points (all framed as documentation structure, never execution state):

  1. planning open (adversarial) — new plan document naming compliance; .dev/plans/ has no orphan or duplicate plan documents; plan prose is free of brand/name residue from prior renames. Adversarial posture: actively look for stale terminology, misplaced knowledge (content that belongs in docs/ stranded in the plan body), and missing plan-hygiene markers.

Read the full file on GitHub · 78 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. 3d ago First seen · 78 lines · 59 tokens per session scan A f7c14510174c

Subscribe to this mod's changes

golem-steward is an agent published in the GitHub repository monkey1wizard/Golem-Agents-Legion (15 stars, last pushed 5d ago), licensed MIT. It adds 59 tokens to every session and 1,897 once invoked, about $0.0003 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 2026-09-04.

Related

Other agents, from other repositories