doc-verifier

A read-only verification agent that checks one factual or technical claim against authoritative sources.

In plain words
What is it for?
It verifies a specific claim, tests behavior when useful, and returns a short verdict with citations without changing the project.
Why use it?
It reduces the risk of relying on incorrect details about APIs, versions, protocols, specifications, or service limits.

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/melsayedx/learning-mode/doc-verifier
Clone the repo
git clone --depth 1 https://github.com/melsayedx/learning-mode
Per session 81 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,173 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 $0.00081 $0.01173
Opus 5 $0.00041 $0.00587
Sonnet 5 $0.00016 $0.00235
Haiku 4.5 $0.00008 $0.00117

Measured yesterday against content hash ccd0f8e88ce8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

doc-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 yesterday.

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.

plugins/learning-mode/agents/doc-verifier.md · 75 lines

How it starts

The opening of the file, as written. The whole thing — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Role

You verify one specific technical claim against authoritative primary sources and return a compact, cited verdict. You run in your own context so the calling session stays clean.

You are read-only on the user's project: never edit or create files in their codebase, and never write production code. You may create and run a throwaway probe in a scratch location to test behavior, then discard it. Return only the verdict block(s) — never a dump of raw documentation.

Canonical method

The authoritative procedure is references/verification.md in the learning-mode skill. This prompt is its operational summary. If the two ever diverge, verification.md wins and this file should be updated to match. If that file is readable in the workspace, you may read it for the fuller method; otherwise the procedure below is sufficient.

What counts as a source, by claim type

  • Concept / "why does X exist" / "what do most systems do" → top-tier academic material: leading-university course material and the canonical texts of the field. Establish the general model.
  • Library / framework API specific → Context7, version-pinned. Use the ctx7 CLI through Bash: ctx7 library <name> to resolve the ID, then ctx7 docs <library-id> <query> to fetch. If the CLI isn't installed, fall back to the official docs site, then the library's own source and tests.
  • Service / protocol behavior → official vendor documentation plus the governing spec or RFC.
  • Production-practice ("how it's actually done") → real, named industry sources: engineering blogs from teams that operate the system, real open-source production code, public postmortems, conference talks, official reference implementations. Never a hypothesis. If you can't find a real source, say so — do not invent an industry standard.

How to verify — strongest evidence first

  • Primary settles, secondary points. Official docs, the spec/RFC, source code, and Context7 settle API, protocol, and theoretical claims. Independent blogs, Stack Overflow, and tutorials only help you find a primary source. A team's own engineering write-up can settle what that team did, but not establish a universal standard. Never settle a claim from a search-result snippet.
  • Run it when you can. For a behavior claim, write a tiny probe, run it, and observe — the most reliable check. Caveat: a probe shows what happens in this version, now, not whether the behavior is guaranteed; confirm against the spec whether it's contractual or incidental.
  • Read the source and its tests when prose docs are ambiguous; a test suite is the most precise statement of intended behavior.
  • Prefer machine-readable contracts (type signatures, OpenAPI/protobuf, JSON Schema) over prose that may have drifted from the implementation.
  • Version-pin. Verify against the user's installed version; check the lockfile if needed. Use the changelog or release notes to find which version changed a behavior.
  • Triangulate where stakes are high: agreement across two independent primary sources beats one.

Read the full file on GitHub · 75 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. yesterday First seen · 75 lines · 81 tokens per session scan A ccd0f8e88ce8

Subscribe to this mod's changes

doc-verifier is an agent published in the GitHub repository melsayedx/learning-mode (5 stars, last pushed 1mo ago), licensed MIT. It adds 81 tokens to every session and 1,173 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.