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 agentmods add instructions/nulljosh/sidewise/claude-mdgit clone --depth 1 https://github.com/nulljosh/sidewiseWhat 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 | $0.01120 | $0.01120 |
| Opus 5 | $0.00560 | $0.00560 |
| Sonnet 5 | $0.00224 | $0.00224 |
| Haiku 4.5 | $0.00112 | $0.00112 |
Grade A, and why
sidewise 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 2d 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 — 55 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Sidewise
RSS news reader across the feeds in src/feeds.js (incl. Hacker News, Fox, BBC, WSJ…). Flat Latest feed + Ground News-style bias view. sidewise.heyitsmejosh.com.
Architecture
Single Cloudflare Worker does everything. worker.js is routing + error handling only; the
logic lives in src/ (feeds parse stories load mcp) so it can be tested without a
Worker runtime. worker.js re-exports all of it, so check-feeds.mjs and tests import one path.
- Fetches every RSS feed in
FEEDS( =[outlet, bias, url, publisher?]insrc/feeds.js; add a source by appending a row — RSS 2.0 or Atom).publisherdefaults tooutletand exists so one newsroom's two feeds (NY Post + NY Post Opinion) count as one voice, not two — counting outlets inflated bias bars and produced false blindspots. Runnpm run feedsafter adding one: it checks recency, not just item count, which is the only way to catch a "zombie" feed that still serves 200 OK from a frozen snapshot (CNN did exactly this for 3 years). parseItemspulls title/link + a timestamp (pubDate/dc:date/published/updated,ts=0when absent).- Returns two views in one
/api/storiesJSON payload:latest— flat reverse-chron across all sources (dateless sinks to bottom), the default reader view.stories—cluster()groups same-story headlines by title-keyword overlap (naive O(n²), seeponytail:comment — upgrade to embeddings if quality matters), tags bias, flags blindspots (covered by one side only).
- Serves the static site (via Workers Static Assets).
public/index.htmlis the marketing page;public/reader.htmlis the actual web app. from the same deploy — no separate Pages project. Frontend defaults to Latest with a source picker + search; tabs switch to the bias view. - The feed pull (not the response) is cached ~2 min via the Cache API under a constant key; responses go out
no-storebecause the zone CDN ignores query strings. A fresh deploy still serves the previous pull until that TTL expires — don't panic if new fields are missing for a couple of minutes. - Failure handling: each feed reports
{ok, items, error}; a pull that loses more than half its feeds isdegradedand falls back to the last good pull (stale: true) rather than serving a half-empty feed, and never overwrites it./api/health503s when degraded.
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.
- 2d ago First seen · 55 lines · 1,120 tokens per session scan A 4992a5f519ad
sidewise CLAUDE.md is an instructions file published in the GitHub repository nulljosh/sidewise (1 stars, last pushed 2d ago), licensed MIT. It adds 1,120 tokens to every session, about $0.0056 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.
Other instructions, from other repositories
clawstash CLAUDE.md
Claude Code instructions for fo0/clawstash, covering claude.md -- project guide, session start -- read order, workflow triggers, output languages and performance / modes.
x402-cloudflare-starter AGENTS.md
AGENTS.md instructions for ANAMIZED/x402-cloudflare-starter, covering agents.md, what this project does, key files, conventions and boundaries.
elc-partnership-builder CLAUDE.md
Claude Code instructions for marian-kamenistak/elc-partnership-builder, covering elc-partnership-builder, what this is, stack, run / build / deploy / test and dev: npm run dev.
gw1-mcp CLAUDE.md
Claude Code instructions for Graphmaxer/gw1-mcp, covering claude.md — gw1-mcp, what this project is, commands, architecture and tech stack.
clawstash AGENTS.md
AGENTS.md instructions for fo0/clawstash, covering agents.md — agent guide and gitnexus — read-only analysis policy (non-negotiable).
toolstop CLAUDE.md
Claude Code instructions for toolstop/toolstop, covering toolstop, adding a server, constraints that are not negotiable, releasing and how the npm tarball is built.