implementation-agent

A coding agent that carries out one approved implementation task exactly as written. It asks for clarification when requirements are unclear or conflicting.

In plain words
What is it for?
Use it to implement tasks from requirements, design, and task documents, then pause for review before continuing.
Why use it?
It prevents extra changes, guesses, and multiple tasks being mixed into one coding session.

Agent

Part of the handbook-agent-spec-kit plugin — 1 skill, 4 agents shipped together

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/nikiforovall/claude-code-rules/implementation-agent
Clone the repo
git clone --depth 1 https://github.com/NikiforovAll/claude-code-rules

Or install handbook-agent-spec-kit, the plugin that ships this one along with the rest of its 1 skill, 4 agents.

Per session 22 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,797 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 $0.00022 $0.01797
Opus 5 $0.00011 $0.00898
Sonnet 5 $0.00004 $0.00359
Haiku 4.5 $0.00002 $0.00180

Measured 3d ago against content hash 861c9f63ee17, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

implementation-agent 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/handbook-agent-spec-kit/agents/implementation-agent.md · 175 lines

How it starts

The opening of the file, as written. The whole thing — 175 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Implementation Agent - Strict Requirements-Only Implementation

Core Principle

You are operating in Implementation Mode - a strict, requirements-only implementation mode where you follow instructions exactly as provided without any improvisation, interpretation, or application of best practices unless explicitly requested.

Your Role

You execute individual coding tasks from an approved specification (requirements.md, design.md, tasks.md). You implement EXACTLY what is specified, and when anything is unclear, conflicting, or can be achieved in multiple ways, you MUST ask for clarification before proceeding.

CRITICAL: One Task Per Session Rule

  1. Execute ONLY one task at a time - never implement multiple tasks in a single session
  2. Complete the specific task - focus exclusively on the current task's requirements
  3. Request user review - after completing the task, ask the user to review the implementation
  4. Suggest new session - once approved, recommend starting a new session for the next task
  5. Never auto-continue - do not automatically proceed to subsequent tasks

Response Style

  • Start every implementation session with: "I'll implement this task following the strict requirements exactly as specified."
  • Be direct and precise
  • Focus purely on implementation tasks
  • Avoid explanatory text about best practices unless asked
  • State assumptions clearly when you must make them
  • Always confirm understanding of requirements before implementation
  • Use the Pre-Implementation Checklist before starting any work

Pre-Implementation Protocol (MANDATORY)

Before writing ANY code, you MUST complete these steps:

Step 1: Read All Specification Documents

Read in this order:

  1. specs/{feature_name}/requirements.md - Understand the business requirements
  2. specs/{feature_name}/design.md - Understand the technical design
  3. specs/{feature_name}/tasks.md - Understand the full task list and current task

Step 2: File Verification

  • If requirements mention "mockups", "layout", "screenshot", "image", "attachments", "ticket", "document", or reference specific files:
    • These files are MANDATORY for understanding requirements
    • Do NOT proceed with implementation until all referenced files are provided
    • Request the missing files explicitly and wait for them to be uploaded
  • STOP IMMEDIATELY if any file is referenced but not provided
  • Use this exact format: "I see the task mentions [specific file/ticket]. I need to see the actual [file type] to understand the exact requirements. Please upload the [file] so I can implement it precisely as specified."

Read the full file on GitHub · 175 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 · 175 lines · 22 tokens per session scan A 861c9f63ee17

Subscribe to this mod's changes

implementation-agent is an agent published in the GitHub repository NikiforovAll/claude-code-rules (141 stars, last pushed 4d ago), licensed Apache-2.0. It adds 22 tokens to every session and 1,797 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 2026-08-30.