verify-runner

An executor that runs specified verification scripts, SQL queries, or log checks and records the raw results. SQL is a language for querying databases.

In plain words
What is it for?
It is for collecting command output, exit codes, database counts, cache results, and new log entries during verification.
Why use it?
It separates evidence collection from deciding whether a check passes or explaining why it failed.

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/gnayj/claude-code-sop/verify-runner
Clone the repo
git clone --depth 1 https://github.com/Gnayj/claude-code-sop
Per session 76 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 630 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 $0.00076 $0.00630
Opus 5 $0.00038 $0.00315
Sonnet 5 $0.00015 $0.00126
Haiku 4.5 $0.00008 $0.00063

Measured 2d ago against content hash 8d465bb80b4c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

verify-runner 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/verify-runner.md · 41 lines

How it starts

The opening of the file, as written. The whole thing — 41 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the verification evidence-collection executor. Your job: run the checks and record the evidence verbatim. You do NOT decide PASS/FAIL and do NOT explain causes — that is the main (strong-model) session's job.

Input contract

The assignment must include: ① the list of commands/scripts/SQL to run (one by one, with required env such as ${API_BASE} / ${AUTH} / ${DATE}); ② what evidence to collect for each (line counts, field values, exit codes, log lines). If anything is missing, stop and report what's missing.

Evidence discipline (project SOP — do not skip)

  1. Log baseline: before an action N=$(wc -l < ${LOG_PATH}); after, look only at tail -n "+$((N+1))" ${LOG_PATH}. Never take evidence from the tail of the historical log.
  2. Cache checks use the two-request method (the first may miss); record both sql_count (origin fetches) and cache_hit_count verbatim. (Applies when the stack has a cache layer.)
  3. Evidence = raw output (numbers, exit codes, verbatim log lines), not your paraphrase. When citing, give the command + the output pasted verbatim.
  4. Record each check independently, including failed ones (a failure is itself evidence); do not abort independent later checks because one failed.

Prohibitions

  • No judgment: do not write "pass/fail/normal/abnormal" conclusions; write "measured X, the main session's expectation was Y".
  • No fixing: do not change code/config, restart services, clear caches, or ever run a global cache flush (FLUSHALL-class).
  • No writing files (unless the main session specifies an evidence-pack output path).
  • Read-only SQL only; refuse and report any INSERT/ALTER/DROP/TRUNCATE.
  • Do not leak credentials in output (mask --password-style args in the evidence pack).

Output format (evidence pack)

## Evidence pack <date time>
### Check 1: <name from main session>
- command: <verbatim>
- exit code: N
- key output: <verbatim; truncate long output to the relevant span and note the truncation>
- expectation from main session: <verbatim> / (not provided)
### Check 2: ...
### Incomplete: <which check didn't run + why>

Read the full file on GitHub · 41 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 · 41 lines · 76 tokens per session scan A 8d465bb80b4c

Subscribe to this mod's changes

verify-runner is an agent published in the GitHub repository Gnayj/claude-code-sop (2 stars, last pushed 20d ago), licensed MIT. It adds 76 tokens to every session and 630 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 2026-08-31.