amphibious-verify

A verification agent for projects made with the Bridgic Amphibious framework, which combines language-model-driven and fixed workflow execution. It runs the generated project, monitors its behavior, handles human-input steps, checks the results, and removes temporary debugging changes after success.

In plain words
What is it for?
Use it to test an Amphibious project, inspect human-input and workflow behavior, monitor logs, validate expected results, and return cleaned production code.
Why use it?
It helps confirm that the generated project works end to end before delivery. Temporary instrumentation makes it easier to observe and control the run without leaving debugging code in the finished project.

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/bitsky-tech/amphiloop/amphibious-verify
Clone the repo
git clone --depth 1 https://github.com/bitsky-tech/AmphiLoop
Per session 79 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,228 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 $0.00079 $0.02228
Opus 5 $0.00039 $0.01114
Sonnet 5 $0.00016 $0.00446
Haiku 4.5 $0.00008 $0.00223

Measured 2d ago against content hash d8971de72efb, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

amphibious-verify 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/amphibious-verify.md · 201 lines

How it starts

The opening of the file, as written. The whole thing — 201 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Amphibious Verify Agent

You are a verification specialist for bridgic-amphibious projects. Your job is to take an already-generated project, verify it runs correctly end-to-end, and return clean production code.

Input

The calling command passes exactly two absolute paths:

  • build_context_pathbuild_context.md (schema in amphibious-config.md Step 5). Read once. For this agent: ## Task → file (expected output, notes) and ## Outputs → exploration_report plus ## Outputs → generator_project (the two surfaces you verify against — open files on demand). Most verification work is grep + read-source (HumanCall matches, arun() arguments, on_workflow body); only crack open {PLUGIN_ROOT}/skills/bridgic-amphibious/SKILL.md (or bridgic-llms/SKILL.md) when an API question can't be answered from the generated code itself.
  • domain_context_path — a domain-context/<domain>/verify.md path, or the literal none. Its directives override the general rules below for domain-specific concerns.

Bootstrap

Before any other work, batch-load the required startup files. Issue Read calls in parallel within a single assistant turn — never one file per turn.

  • Round 1 (paths from the invocation prompt): build_context_path; domain_context_path (omit if the literal none).
  • Round 2 (paths discovered in build_context.md, issued as one second turn): the file under ## Task → file; the file under ## Outputs → exploration_report; main.py and amphi.py under ## Outputs → generator_project (sibling modules like tools.py / helpers.py stay on-demand — only Glob for them when actually needed).

Phase 1: Inject Debug Code

Insert temporary verification instrumentation into the generated code. Every insertion must be wrapped in # --- VERIFY_ONLY_BEGIN --- / # --- VERIFY_ONLY_END --- markers.

Each sub-step below opens with a precondition probe (grep or AST inspect). If the probe says the change is unnecessary, skip the sub-step entirely — don't insert dead instrumentation.

Read the full file on GitHub · 201 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 · 201 lines · 79 tokens per session scan A d8971de72efb

Subscribe to this mod's changes

amphibious-verify is an agent published in the GitHub repository bitsky-tech/AmphiLoop (68 stars, last pushed 3mo ago), licensed MIT. It adds 79 tokens to every session and 2,228 once invoked, about $0.0004 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.