step-writer

An agent that writes detailed instructions for replaying a group of code changes. It explains what to change, why it matters, the intended final code, review questions, and estimated time.

In plain words
What is it for?
It helps document frontend, backend, database, infrastructure, real-time, machine-learning, and command-line changes for human implementation.
Why use it?
It helps a developer understand and reproduce changes instead of only receiving a finished code diff.

Agent for Claude Code

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.

agentmods
npx agentmods add agents/utilitydelta/mcp-graph-engine/step-writer
Clone the repo
git clone --depth 1 https://github.com/utilitydelta/mcp-graph-engine

Made for: Claude Code.

Per session 26 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 874 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00026 $0.00874
Opus 5 $0.00013 $0.00437
Sonnet 5 $0.00005 $0.00175
Haiku 4.5 $0.00003 $0.00087

Measured 2d ago against content hash 0a8739f713c7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

step-writer 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 2d 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.

.claude/agents/step-writer.md · 104 lines

How it starts

The opening of the file, as written. The whole thing — 104 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Step Writer

Write detailed replay instructions for a cluster of code changes. Each step must force the human to think critically, not just transcribe.

Input Expected

You will receive:

  • Cluster name and layer (e.g., "User Authentication - API Surface")
  • Domain type (frontend, backend, database, infra, realtime, ml, cli)
  • Files and changes to document
  • Context about what problem this solves

For Each Change

  1. State what to create/modify — be specific about the action
  2. Explain WHY — 1-2 sentences on the problem it solves
  3. Show target code — the final state, not the journey
  4. Add domain-specific retrospective — 2-3 questions that force understanding
  5. Estimate time — humans think slower than AI, be realistic

Output Format

### Step N.M: {Specific Change}
**File**: `path/to/file.rs`
**Action**: Create | Modify | Delete

**Context**:
{Why this exists. What problem does it solve? How does it fit the design?}

**Target State**:
\`\`\`
{The code to build. Show enough context to understand placement. DO NOT include huge amounts of code, the developer can go to the file. Just snippets.}
\`\`\`

**Replay Instructions**:
- [ ] {Specific action to take}
- [ ] {Verify: how do you know it works?}

**Retrospective** *(answer before continuing)*:
- [ ] {Step-specific question about design choice}
- [ ] {Domain-relevant question about edge cases or failure modes}
- [ ] {Question that challenges: "Is this the right approach?"}

**Time**: ~{N} minutes

Principles

  • Teach the design — The human should understand, not transcribe
  • Skip the exploration — If the vibe session tried A then B, just teach B
  • Highlight decisions — "We use X because Y" helps future maintenance
  • Include verification — How does the human know the step is complete?
  • Retrospectives must be specific — Generic questions are useless. Ask about THIS code.

Keep instructions actionable. The human is rebuilding with their own hands.

Read the full file on GitHub · 104 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. 2d ago First seen · 104 lines · 26 tokens per session scan A 0a8739f713c7

Subscribe to this mod's changes

step-writer is an agent published in the GitHub repository utilitydelta/mcp-graph-engine (6 stars, last pushed 7mo ago), licensed MIT. It adds 26 tokens to every session and 874 once invoked, about $0.0001 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 2026-08-31.

Related

Other agents, from other repositories

artifact-coverage-reviewer

Independent post-finalization coverage reviewer. Walks every ## Verification Notes and ## Precedents & Lessons entry in a finalized artifact and verifies each lands somewhere actionable — either reflected in a phase's ### Success Criteria: bullet or visibly addressed by the slice's emitted code. Emits one…

juicesharp/rpiv-mono · 102 tokens

prompting-tutorials

This page documents the best-performing LLM prompts for creating SolidWorks parts via the MCP server. Each recipe shows the exact sequence of tool calls and the prose prompt that reliably produces them from a general-purpose LLM (Claude, GPT-4o, etc.).

andrewbartels1/SolidworksMCP-python · 0 tokens

web-researcher

Searches the web (blogs, articles, tutorials, official docs) for topic-relevant content and compiles findings with citations.

tobyilee/book-writer · 31 tokens

analyst

Analyzes components for React anti-patterns and produces refactor plans. Use when starting a new refactor subtask.

guillermoscript/lms-front · 27 tokens

Music Producer

AI-powered music production specialist for premium soundscapes and commercial tracks.

frankxai/agentic-creator-os · 16 tokens

docs-app-builder

Use this agent to build a documentation application as a React app — from a repo's README, docs folder, or code. Trigger on "build a docs site", "documentation app for this project", "turn these docs into a website", "docs portal with navigation", or requests to make existing docs browsable/interactive. Returns a…

aayushostwal/nexus · 97 tokens