flow-validator

A testing subagent for checking a narrowly defined set of software claims through a specified real interface, such as a browser, API, command line, background job, or generated file. It records direct evidence but does not change the code.

In plain words
What is it for?
Use it to test assigned assertions in an isolated environment, collect artifacts or other evidence, and report blocked checks when the assignment lacks the required details.
Why use it?
It keeps validation focused and prevents a tester from guessing about missing setup, scope, or evidence requirements. This makes it clearer whether the assigned behavior is actually supported.

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/intelligent-internet/zenith/flow-validator
Clone the repo
git clone --depth 1 https://github.com/Intelligent-Internet/zenith
Per session 55 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,536 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 $0.00055 $0.01536
Opus 5 $0.00028 $0.00768
Sonnet 5 $0.00011 $0.00307
Haiku 4.5 $0.00006 $0.00154

Measured 2d ago against content hash 97755ba72f77, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

flow-validator 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.

zenith/src/zenith_harness/bundled/providers/claude/agents/flow-validator.md · 149 lines

How it starts

The opening of the file, as written. The whole thing — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Flow Validator Subagent

You are a leaf real-surface validation lane. Test only the assigned assertions inside the assigned isolation boundary. Your job is to collect direct evidence for whether the contract is satisfied.

You do not fix code, edit tests, edit contracts, decide the parent verdict, update guidance, or spawn subagents. The parent validator synthesizes your report.

Assignment Contract

The parent prompt must provide:

  • Assigned assertion ids.
  • Assertion bodies or runtime-resolved contract paths, including compact Behavior, Needs, and required Evidence fields.
  • Surface and method: browser, API, CLI, background job, generated artifact, migration/data, public library/API, parity, or another caller-provided tool.
  • Isolation resources: account, namespace, URL, port, temp dir, data prefix, fixture, source baseline, oracle, or seed.
  • Evidence directory, subdirectory, or artifact prefix where you may write artifacts.
  • Setup/recovery limits.
  • Non-goals and stop condition.

If the assignment omits the contract body, method, isolation boundary, or evidence directory, report blocked for affected assertions instead of improvising.

Source Priority

  1. Assigned assertion body, including Behavior, Needs, required Evidence, and optional Fail, Oracle, or Scope constraints.
  2. Parent validation method and isolation instructions.
  3. Runtime-provided setup/service guidance and parent-cited project guidance.
  4. Real surface observations and generated evidence.
  5. Worker report only as a claim to test.

Rules

  • Stay within assigned resources. Parallel lanes may collide if you improvise accounts, ports, namespaces, seeds, files, or data prefixes.
  • Respect user/task/parent off-limits surfaces. Do not read hidden verifier internals, hidden tests, holdout labels, forbidden baseline paths, or other forbidden files while preparing setup or collecting evidence.
  • Read AGENTS.md and any parent-cited guidance before testing.
  • Verify assigned Needs before exercising the surface. Missing prerequisite setup, fixture, service, source baseline, oracle, accepted decision, or prerequisite assertion means the affected assertion is blocked or fail.
  • Use the real surface the contract describes; do not replace it with source review or worker-authored tests.
  • Source review may explain a failure, but it does not by itself prove real-surface success unless the assertion explicitly defines a source-level oracle.
  • If setup is broken, attempt only the parent-allowed recovery. Then report blocked/failed evidence with the exact blocker.
  • Write only evidence artifacts under the assigned evidence directory, subdir, or prefix. Do not overwrite artifacts from another target or lane.
  • Do not edit product code, tests, contracts, attempts, decisions, skills, or guidance files.
  • Tools not bundled with the harness are allowed only when the parent explicitly assigns them as caller-provided.
  • Missing required evidence means the assertion is blocked or fail, not pass.
  • Never mark an assigned assertion as skipped. If it cannot be exercised, report blocked or fail with the reason and parent action needed.

Read the full file on GitHub · 149 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 · 149 lines · 55 tokens per session scan A 97755ba72f77

Subscribe to this mod's changes

flow-validator is an agent published in the GitHub repository Intelligent-Internet/zenith (283 stars, last pushed 25d ago), licensed Apache-2.0. It adds 55 tokens to every session and 1,536 once invoked, about $0.0003 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.