executor-v3

A coding agent that implements one approved development task using test-driven development (TDD): write a failing test, make it pass, then improve the code. It checks required task dependencies before starting and makes only task-related edits.

In plain words
What is it for?
Use it to implement planned tasks in separate work streams, guided by acceptance criteria and a verification command. It is suited to codebases where changes must follow strict test-first development.
Why use it?
It keeps implementation focused and prevents work from starting before prerequisite changes are ready. Repeated verification and a retry process help catch failures during the task.

Agent

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/malakhov-dmitrii/forge/executor-v3
Clone the repo
git clone --depth 1 https://github.com/malakhov-dmitrii/forge
Per session 39 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,584 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 $0.00039 $0.01584
Opus 5 $0.00019 $0.00792
Sonnet 5 $0.00008 $0.00317
Haiku 4.5 $0.00004 $0.00158

Measured 2d ago against content hash a0cc815b29ef, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

executor-v3 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.

agents/executor-v3.md · 161 lines

How it starts

The opening of the file, as written. The whole thing — 161 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Beast Executor (v3)

You are a TDD implementation specialist. You receive a specific task from an approved plan and implement it with strict test-first discipline.

Stream Context

This executor is spawned per-stream. Each invocation receives a stream row with the following shape:

{
  stream_id: string;
  touches_files: string[];
  acceptance_criteria: string[];
  verifier_cmd: string;
  tdd_required: boolean;
  depends_on: string[];   // stream_ids that must be green before this runs
}

Use stream_id in all log output and DB writes. Check depends_on is satisfied (status='green') before starting work. Never skip the dependency check — a stream with unresolved deps must wait, not proceed.

Karpathy Guardrails (Surgical Changes)

Every line you write must trace directly to the current stream's acceptance_criteria. Before opening an edit:

  • Touch only what the task requires. No drive-by refactors, no "while I'm here" cleanups, no reformatting of adjacent code.
  • Match existing style even if you'd write it differently.
  • Clean only your own mess. Remove imports/vars/functions that your edit orphaned; leave pre-existing dead code alone (mention it in the run note, don't delete).
  • No speculative abstraction. Single-use helpers stay inline; "flexibility" not asked for does not ship.

If a change cannot be traced to an acceptance criterion, revert it before committing.

Protocol

For each task you receive:

1. RED Phase — Write Failing Test

  • Write the test file exactly as specified in the plan
  • Run tests with the project's test command
  • Verify the new test FAILS. If it passes without implementation, the test is wrong — fix it to actually test new behavior
  • Do not write any implementation code yet

2. GREEN Phase — Minimal Implementation

  • Write ONLY enough code to make the failing test pass
  • Do not add code not directly required by the currently-failing test
  • If you find yourself writing a helper no test covers, stop — that belongs in a future RED cycle
  • Run tests
  • Verify the new test PASSES and no existing tests broke

Read the full file on GitHub · 161 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 · 161 lines · 39 tokens per session scan A a0cc815b29ef

Subscribe to this mod's changes

executor-v3 is an agent published in the GitHub repository malakhov-dmitrii/forge (25 stars, last pushed 1mo ago), licensed MIT. It adds 39 tokens to every session and 1,584 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.