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.
npx skills add The-Artificer-of-Ciphers-LLC/skills-from-the-artificer --skill trust-but-verifygit clone --depth 1 https://github.com/The-Artificer-of-Ciphers-LLC/skills-from-the-artificerWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/the-artificer-of-ciphers-llc/skills-from-the-artificer/trust-but-verify)<a href="https://agentmods.dev/skills/the-artificer-of-ciphers-llc/skills-from-the-artificer/trust-but-verify"><img src="https://agentmods.dev/badge/skills/the-artificer-of-ciphers-llc/skills-from-the-artificer/trust-but-verify/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/the-artificer-of-ciphers-llc/skills-from-the-artificer/trust-but-verify"><img src="https://agentmods.dev/badge/skills/the-artificer-of-ciphers-llc/skills-from-the-artificer/trust-but-verify.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00176 | $0.02638 |
| Opus 5.5 | $0.00070 | $0.01055 |
| Sonnet 5.5 | $0.00035 | $0.00528 |
| Haiku 4.5 | $0.00018 | $0.00264 |
Grade A, and why
trust-but-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 1mo 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.
How it starts
The opening of the file, as written. The whole thing — 107 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Trust but Verify
A subagent's report is a lead, not a fact. You trust it enough to investigate — never enough to act on unread.
When you delegate work to a subagent, what comes back is a claim about reality, filtered through a model that summarized, paraphrased, and sometimes hallucinated. The agent may be right. It may be confidently wrong. You cannot tell which from the report alone — the wrong ones look exactly as fluent as the right ones.
This skill is the discipline of closing that gap: before you build on, repeat, or report any claim an agent handed you, you re-open the primary source and confirm the claim against it. Every claim. With a citation.
The one rule everything else serves
No claim is verified until you have opened a primary source that confirms it and can quote where.
Two corollaries that do the real work:
- Inference is not verification. "That's how this library usually works", "the function name implies it returns null", "that lines up with what I'd expect" — these are guesses wearing a lab coat. They produce an
UNVERIFIED, never aVERIFIED. - The agent's own words are not a source. "The agent said the timeout is 30s" verifies nothing. The agent is the thing under audit. The source is the config file where
30sis written.
If you cannot find a source, the honest verdict is UNVERIFIED — and you say so out loud. Silently dropping an unsourced claim is worse than flagging it, because a clean report reads as "all confirmed."
The loop
For a returned report, work claim by claim:
-
Atomize. Break the report into discrete, individually-checkable assertions. "The retry logic is in
cleanup.ts, runs 3 times, and was added in #493" is three claims (location, count, provenance), each with its own source and its own verdict. Bundled claims hide the false one inside the true ones. -
Classify → pick the source. Each claim has a type, and each type has an authoritative source (see the ladder below). A claim about code is settled by the code, not by docs about the code. A claim about a library's behavior is settled by that library's docs or source, not by the codebase that calls it.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago First seen · 107 lines · 176 tokens per session scan A acf8709e1d48
trust-but-verify is a skill published in the GitHub repository The-Artificer-of-Ciphers-LLC/skills-from-the-artificer (6 stars, last pushed yesterday), licensed MIT. It adds 176 tokens to every session and 2,638 once invoked, about $0.0007 per session on Opus 5.5. 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.
Other skills, from other repositories
incident-postmortem
Escribe un postmortem de incidente estructurado o revisión post-incidente. Úsalo cuando se te pida escribir un postmortem, reporte de incidente, revisión P1/P2, reporte de apagón o RCA (análisis de causa raíz). Produce un postmortem sin culpables con cronología, causa raíz, factores contribuyentes, resumen de impacto…
agent-incident-postmortem
Run a blameless postmortem for an incident caused by an AI agent or LLM feature — hallucinated facts shipped to users, runaway tool use, prompt injection, cost blowouts, or wrong actions taken autonomously. Use when asked to write up an AI incident, analyse why an agent did something wrong, or produce corrective…
agent-design-review
Review an LLM agent design and find where it will be unreliable, expensive, or unsafe. Use when asked to review an agent architecture, critique a multi-step/tool-using agent, debug an agent that loops or goes off-task, or harden an agent before launch. Produces a structured review — task fit, control flow, tools…
prompt-debugging
Figure out why a prompt isn't working and fix it — diagnose the actual failure (ambiguity, missing context, wrong format, conflicting instructions) instead of randomly rewording. Use when asked why isn't my prompt working, the AI keeps ignoring my instructions, my prompt gives inconsistent results, or how do I fix…
prompt-optimizer
Diagnose and rewrite an underperforming LLM prompt so it produces reliable, well-structured output. Use when asked to improve a prompt, fix a prompt that gives inconsistent or wrong results, reduce hallucination/refusals, or make output follow a format. Produces a rewritten prompt with a diagnosis of what was failing…
rag-architecture-review
Review an existing Retrieval-Augmented Generation system and find why it underperforms. Use when asked to review or audit a RAG pipeline, diagnose wrong/ungrounded answers from a 'chat with your docs' feature, or improve an already-built knowledge assistant. Produces a staged review — ingestion, chunking, retrieval…