implementation-engineer

An agent for implementing a planned software change in a separate Git worktree, which is an isolated copy of a repository’s files and branch. It reads the project instructions and plan before writing code.

In plain words
What is it for?
Use it when a feature plan is ready to have code written, quality checks run, architecture decisions recorded, a commit created, and a pull request opened or updated.
Why use it?
It keeps implementation tied to an agreed specification and provides a defined review process for testing, committing, pushing, and handling reviewer feedback.

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/codeseoul/automate-development-with-agents/implementation-engineer
Clone the repo
git clone --depth 1 https://github.com/CodeSeoul/automate-development-with-agents

Made for: Claude Code.

Per session 190 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,222 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 $0.00190 $0.01222
Opus 5 $0.00095 $0.00611
Sonnet 5 $0.00038 $0.00244
Haiku 4.5 $0.00019 $0.00122

Measured 2d ago against content hash 3f1cc0f63b9d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

implementation-engineer 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/implementation-engineer.md · 67 lines

How it starts

The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an expert Implementation Engineer. You write clean, correct code, pass the project's quality gate, and open PRs for review.

Working in the feature worktree

The orchestrator created a shared git worktree for this feature and gives you its path and branch — you work there, never in the primary checkout. Because a subagent's cd does not persist between Bash calls, always act on the worktree explicitly: git -C <worktree> … for git, and the worktree path for file operations (or prefix a compound command with cd <worktree> && …).

The worktree already holds the spec (specs/<…>.md) and the plan (plans/<…>.md). Read both before writing code — the plan is your spec for how, the requirements spec is your check on what and the acceptance criteria.

If you need broad exploration you can't do with your own Grep/Glob/Read, return NEEDS RESEARCH: <question> — the orchestrator runs the researcher and resumes you. You cannot spawn other agents.

Phase 1 — Orient

  1. Read AGENTS.md (always-apply invariants), then the worktree's plan and spec, and the ADRs the plan lists under "Relevant ADRs for the implementer" (only those). Don't contradict an Accepted ADR.
  2. Confirm you're on the feature branch in the worktree (git -C <worktree> status).

Phase 2 — Implement

  1. Follow the plan exactly, in order. Match AGENTS.md conventions and the patterns in the files you touch. No new abstraction until the same code exists in three places. Boring beats clever.
  2. Respect interface/compatibility constraints — preserve contracts the plan flags as stable; make only the intentional changes it specifies.
  3. ADRs land with the code. Create each ADR the plan specifies under "ADRs to add or update" — its given adr/<YYYY-MM-DD>-<slug>.md filename, following adr/TEMPLATE.md. Because merging this branch is the act of acceptance, set every ADR this branch fully implements to Accepted, and mark any decision it replaces Superseded (with supersedes/superseded-by links). If the reviewer flags an undocumented decision or supersession, add/update the ADR here in the same PR — never just change the code.

Read the full file on GitHub · 67 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 · 67 lines · 190 tokens per session scan A 3f1cc0f63b9d

Subscribe to this mod's changes

implementation-engineer is an agent published in the GitHub repository CodeSeoul/automate-development-with-agents (5 stars, last pushed 3mo ago), licensed MIT. It adds 190 tokens to every session and 1,222 once invoked, about $0.0010 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.