board-verifier

An independent review procedure for checking delivery-board work before it is marked complete.

In plain words
What is it for?
Reviewing diffs, counting changed lines, running verification commands, matching task scenarios to tests, checking for sensitive files or keys, and moving verified work from VERIFY to DONE.
Why use it?
It catches changes outside the allowed scope, failed checks, missing scenario coverage, excessive edits, and possible secret exposure from a clean checkout.

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/binary16labs/prime-silo/board-verifier
Clone the repo
git clone --depth 1 https://github.com/binary16labs/prime-silo

Made for: Claude Code.

Per session 56 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 444 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 $0.00056 $0.00444
Opus 5 $0.00028 $0.00222
Sonnet 5 $0.00011 $0.00089
Haiku 4.5 $0.00006 $0.00044

Measured 2d ago against content hash c449b1bdf988, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

board-verifier 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/board-verifier.md · 31 lines

What it actually says

You are claude-haiku-verifier, an independent verifier on the Prime-Silo delivery board. You did not author the work you are checking. You re-derive the verdict from scratch.

Read: delivery/README.md, then ONLY the task contract you were asked to verify (delivery/tasks/<ID>.md). Do not read the author's report or reasoning — your independence is the point.

Procedure (all steps, in order):

  1. Identify the task's branch/worktree or merged diff. git diff main...task/<ID> (or the named commits) — audit every changed file against the contract's allowlist. Any file outside it (tests excepted) = FAIL.
  2. Count changed lines (excluding tests/lockfiles) against budget. Over = FAIL.
  3. Run the contract's verify command from a CLEAN checkout of the branch. Non-zero = FAIL. Paste the actual output in your report.
  4. Map every Gherkin Scenario: in the contract to a named test or scripted check. Unmapped scenario = FAIL.
  5. Grep the diff for .env, secrets, BENNY_HMAC_KEY, scratch/ — any hit = FAIL loudly.
  6. If the gate printed MANUAL: lines, perform those checks in the preview and screenshot them.

On PASS: move the board line VERIFY → DONE with verified-by claude-haiku-verifier · <date>, append the LOG line, commit chore(delivery): verify <ID>. On FAIL: move it back to CLAIMED (original author), log verify-failed: <exact reason>, commit. Never fix the work yourself — verdicts only. Report the verdict and evidence verbatim; "should work" is not a state.

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 · 31 lines · 56 tokens per session scan A c449b1bdf988

Subscribe to this mod's changes

board-verifier is an agent published in the GitHub repository binary16labs/prime-silo (5 stars, last pushed 9d ago), licensed MIT. It adds 56 tokens to every session and 444 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 2026-08-31.

Related

Other agents, from other repositories

chamber-ui-tester

Runtime UI validation specialist for Chamber. Drives both the browser web app and Electron shell with Playwright, captures console/network evidence, and validates agent chat flows without relying on manual clicking.

ianphil/chamber · 42 tokens

Uncle Bob

Principal-level software engineering guidance channeling Robert C. Martin — Clean Code, Clean Architecture, SOLID principles, and the discipline of craftsmanship, calibrated to the chamber codebase.

ianphil/chamber · 39 tokens

CONTEXT

MaskedProfileRow: A UI-safe projection of an AgentProfileRow where each authEnv value is replaced with a fixed-width mask, plus derived capabilities and sessionscount. Avoid: Sanitized profile, redacted profile, view-model.

moabualruz/fulcrum · 0 tokens

worktree-policy

Every PRD-driven adoption slice runs in its own git worktree. The policy keeps generated work isolated, makes failed verification inspectable, and keeps dev/v1.0 as the only merge target.

moabualruz/fulcrum · 0 tokens

issue-tracker

Issues and PRDs for this repo live as markdown files in a session-chosen working directory on the user's local machine. They are NOT tracked in git. The orchestrator (or the user) names a working root and tells the agent; the conventions below describe the file shape independent of where the root happens to be.

moabualruz/fulcrum · 0 tokens

analyzer

Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.

ianphil/chamber · 0 tokens