prime-silo CLAUDE.md

prime-silo CLAUDE.md is an instructions file for coding agents from binary16labs/prime-silo. It costs 4,710 tokens per session, scanned A, original, MIT.

A set of instructions for agents working in the Prime-Silo repository, including its current status, architecture, navigation guidance, and authority boundaries. It describes a project that includes a website and data-lineage features.

In plain words
What is it for?
Use it when navigating Prime-Silo, checking its current state, or deciding what agents are allowed to change.
Why use it?
It gives agents the project context and limits they need before making changes in Prime-Silo.

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/binary16labs/prime-silo/claude-md
Clone the repo
git clone --depth 1 https://github.com/binary16labs/prime-silo

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 prime-silo CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/binary16labs/prime-silo/claude-md.svg)](https://agentmods.dev/instructions/binary16labs/prime-silo/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/binary16labs/prime-silo/claude-md"><img src="https://agentmods.dev/badge/instructions/binary16labs/prime-silo/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 4,710 This file is loaded in full into every session.
When invoked 4,710 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 $0.04710 $0.04710
Opus 5 $0.02355 $0.02355
Sonnet 5 $0.00942 $0.00942
Haiku 4.5 $0.00471 $0.00471

Measured 3d ago against content hash 1f7bd579a19e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

prime-silo 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 3d 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 · 439 lines

How it starts

The opening of the file, as written. The whole thing — 439 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Prime-Silo Agent Navigation

For Claude and other agents working on Prime-Silo


Current Status (as of 2026-07-16)

Latest Release: v1.16.2 (marketing site: LONGVIEW coverage proof + additive lineage-mesh DAG governance anchor)
New in 1.16.2 — website (website/, auto-deploys to GitHub Pages on main): turned v1.16.x work into numbered proof — LONGVIEW 46.3% arc-driven coverage stat (from 31.4% baseline), OpenLineage lineage mesh + Sovereign Memory Teleport in governance/features, and a new scroll-scrubbed lineage-mesh DAG chapter (#lineage, additive alongside the seal chapter) whose topology mirrors dashboard/lineage.mjs. See the animejs-scrollcraft skill for the chassis + hard-won gotchas.
New in 1.16.x — governance & privacy layer over LONGVIEW:

  • Memory management (teleport) — every LONGVIEW artifact is sid-keyed, so a label over project/title/card-text resolves deterministically to sessions, and scripts/longview/memory.mjs teleports their whole lineage subtree (cards/evidence/windows/reviews + Neo4j Sources + exclusive Concepts + Chroma chunks, via memory_graph.py) into a quarantine workspace — moves, never deletes; journalled, fully reversible (restore). longview/quarantine.json filters teleported sids from every future inventory read. A deterministic leak gate (lib/leak_gate.mjs) scans deliverables for terms+sids before sharing. Search/teleport UI at the dashboard :8788/memory.html.
  • Observability dashboard (scratch/longview_run/dashboard/, bash dash.sh:8788) — deterministic, ledger-sourced (single source of truth). Mission Control (/): pipeline lineage rail with done/total denominators, live graph stats, enrich stage panes, ontology, ES5 Kindle page (/kindle.html). Lineage & governance (/lineage.html): OpenLineage-format DAG (lineage.mjs → RunEvents in longview/lineage/openlineage.json), click-to-drill artifact explorer (full Section→Evidence→Cards→Windows→Sessions→Concepts lineage + step-through + ontology, reusing lib/record.mjs — the same disk-truth as the app's benny_record), and a governance execution register (every run: commit/model/phases/outcome/cost) groupable by output-ontology/workspace.
  • Arc-driven bookopus now plans timeline-walked cross-project arcs (lib/arcs.mjs: threads that span ≥2 sessions/≥2 projects, narrated from real beats) and writes sections draft → gate → connective-critique → revise. LONGVIEW_OPUS_DIR builds an iteration into data_out/iterations/<name>/ without clobbering a prior book.
  • Wedge hardening — stall watchdog (not wall-clock), per-call LLM deadlines + circuit breaker, subprocessEnv forwards .env to spawned python, reasoning_effort:none fix (empty-content fallback), and a rag ingest cancel endpoint + circuit breaker (POST /api/rag/ingest/cancel, GET /api/rag/ingest/active) so a dead client can't flood the LM host for hours.
  • Coverage bug fixed — the opus coverage denominator counted .meta.json (both end in .json); baseline book was reported 15.7% but is truly 31.4% (59/188).

Read the full file on GitHub · 439 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. 3d ago First seen · 439 lines · 4,710 tokens per session scan A 1f7bd579a19e

Subscribe to this mod's changes

prime-silo CLAUDE.md is an instructions file published in the GitHub repository binary16labs/prime-silo (5 stars, last pushed 10d ago), licensed MIT. It adds 4,710 tokens to every session, about $0.0235 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.