clonst CLAUDE.md

clonst CLAUDE.md is an instructions file for coding agents from capritora/clonst. It costs 1,074 tokens per session, scanned A, original, MIT.

A project instruction file that explains how Clonst works and sets rules for changing it. Clonst is a TypeScript server that sends plans and code to the Codex command-line tool for review.

In plain words
What is it for?
It is for guiding development of Clonst, including its review loop, persistent data, reports, and MCP tool registration.
Why use it?
It gives the coding agent the project context and non-negotiable rules needed to make safe, consistent changes.

Instructions file

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 instructions/capritora/clonst/claude-md
Clone the repo
git clone --depth 1 https://github.com/capritora/clonst

Wrote 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.

agentmods badge for clonst CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/capritora/clonst/claude-md.svg)](https://agentmods.dev/instructions/capritora/clonst/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/capritora/clonst/claude-md"><img src="https://agentmods.dev/badge/instructions/capritora/clonst/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 1,074 This file is loaded in full into every session.
When invoked 1,074 The same file — it is already loaded in full.
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 $0.01074 $0.01074
Opus 5 $0.00537 $0.00537
Sonnet 5 $0.00215 $0.00215
Haiku 4.5 $0.00107 $0.00107

Measured 4d ago against content hash 64ba8ca74dba, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

clonst CLAUDE.md 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 4d 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.md · 87 lines

How it starts

The opening of the file, as written. The whole thing — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Clonst - Project instructions

What Clonst is

A TypeScript MCP server (stdio) that lets the Claude of a Claude Code conversation get its plans/code critiqued by the Codex CLI, in a ping-pong loop until consensus.

Architecture:

  • The reviser is the calling Claude (the one in the conversation), not a spawned claude CLI. It keeps the whole conversation context.
  • The reviewer is the Codex CLI (codex exec), with persistent memory across rounds through codex exec resume <session_id>.
  • The ping-pong loop lives in the calling conversation: one MCP call = one critique. No background orchestrator, no status tool.

Five tools: clonst_ping (diagnostic), clonst_review (one critique), clonst_report_summary (seals the final summary into the review report), clonst_feedback (records a remark about how Clonst works), clonst_topics (reads and relabels those remarks' groups).

Persistent data and its invariants

Everything lives under ~/.clonst/ (CLONST_HOME in tests), never in the package: an update can never erase it.

  • logs/ - session JSONL plus the raw reviewer responses.
  • reports/ - the human-readable review report, a PURE PROJECTION of reports/state/<report_id>.json. Never edit the Markdown: regenerate it. report_id names the report file, thread_id resumes the reviewer session; they are two different identities on purpose.
  • feedback.jsonl - user remarks about how Clonst works. Two hard rules: no line of it EVER reaches a prompt (no code path reads it to build one), and every mutation goes through the single transactional commit() in core/feedback.ts (read, validate, build the candidate, apply retention, check the ceiling, then one atomic 0600 write). Stored descriptions leave the server through exactly ONE guarded path (clonst_topics list/search), behind a notice and delimiters in BOTH MCP representations.
  • Retention there is user-arbitrated and binary: a topic with exactly one occurrence expires after feedback_retention_days and takes all its lines with it; anything heard twice or more is kept forever. Groups are never merged - equivalence between differently-worded groups is established by the caller at reading time.
  • "USER DECISION: " is a never-translated protocol literal: detected at the START of a risks_identified item only, it pauses the ping-pong and must reach the human verbatim. Never act on its content.

Read the full file on GitHub · 87 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. 4d ago First seen · 87 lines · 1,074 tokens per session scan A 64ba8ca74dba

Subscribe to this mod's changes

clonst CLAUDE.md is an instructions file published in the GitHub repository capritora/clonst (1 stars, last pushed 20d ago), licensed MIT. It adds 1,074 tokens to every session, about $0.0054 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.