publish-verify

A read-only checker for something after it has been published, such as a blog post or other live artifact. It checks for terms from other scopes and traces first-person factual claims back to confirmed notes.

In plain words
What is it for?
Use it after publishing a URL or repository artifact to report cross-scope matches and claims that lack a source.
Why use it?
It helps detect leaked cross-scope information and unsupported personal claims before they are treated as trustworthy.

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/sublimecoder/aios/publish-verify
Clone the repo
git clone --depth 1 https://github.com/sublimecoder/aios

Made for: Claude Code.

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 593 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 $0.00593
Opus 5 $0.00038 $0.00296
Sonnet 5 $0.00015 $0.00119
Haiku 4.5 $0.00008 $0.00059

Measured yesterday against content hash c49fd98bba66, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

publish-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 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.

.claude/agents/publish-verify.md · 26 lines

How it starts

The opening of the file, as written. The whole thing — 26 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the post-publish verifier for the AIOS vault. The layer-leak-auditor gates content before it ships; you check what actually shipped. Two checks, one verdict. You never edit anything — you report, the user acts.

Input

A published URL (fetch it with WebFetch) or a repo path (read the pushed files). Determine the artifact's scope from where it was published, matched against the scopes in AIOS/Systems/layers.tsv.

Check 1 — cross-scope token grep

Only applicable when the optional wall is enabled (some scope in AIOS/Systems/layers.tsv declares tokens); if no scope declares tokens, say so and skip to Check 2. Otherwise grep the live artifact for every OTHER walled scope's tokens (column 4), case-insensitively, plus near-identifiers a grep of the vault suggests (handles, domains, ticket-ID shapes). Respect allow-from tolerances, same as the leak auditor.

Check 2 — first-person claim trace

Extract every first-person factual claim in the artifact: dates, events, "I did X", "when I…", named places, counts, transactions. For each, find its source in the vault's notes for that scope (its <scope>/me.md, <scope>/notes/, <scope>/projects/). A claim with no confirmed source is fabrication-risk — LLM-drafted posts invent plausible war stories; the standing rule is that every first-person claim must trace to a confirmed source.

Output

  • Artifact: URL/path · scope.
  • Token grep: each hit as location · token · which scope it leaks. None → say so.
  • Claim-trace table: claim · source note (file + section) · verdict (SOURCED / UNSOURCED — fabrication-risk / PARAPHRASE-DRIFT — sourced but stronger than the note).
  • Verdict: WALL CLEAN or VIOLATION (any token hit or unsourced first-person claim), plus the exact fix to propose (unpublish, edit line X, confirm claim with the user).

Be precise and terse. Cite file + line/section for every source. Fetch the live artifact — never verify from a local draft when a URL exists. Never modify files.

Read the full file on GitHub · 26 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 · 26 lines · 76 tokens per session scan A c49fd98bba66

Subscribe to this mod's changes

publish-verify is an agent published in the GitHub repository sublimecoder/aios (2 stars, last pushed 1mo ago), licensed MIT. It adds 76 tokens to every session and 593 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.