Borrowing it
Nothing to install: this file belongs to DataSift-Ty-Personal/SiftStack. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/DataSift-Ty-Personal/SiftStack/main/CLAUDE.mdgit clone --depth 1 https://github.com/DataSift-Ty-Personal/SiftStackWrote 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.
[](https://agentmods.dev/instructions/datasift-ty-personal/siftstack/claude-md)<a href="https://agentmods.dev/instructions/datasift-ty-personal/siftstack/claude-md"><img src="https://agentmods.dev/badge/instructions/datasift-ty-personal/siftstack/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/datasift-ty-personal/siftstack/claude-md"><img src="https://agentmods.dev/badge/instructions/datasift-ty-personal/siftstack/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.20414 | $0.20414 |
| Opus 5.5 | $0.08166 | $0.08166 |
| Sonnet 5.5 | $0.04083 | $0.04083 |
| Haiku 4.5 | $0.02041 | $0.02041 |
Grade A, and why
SiftStack 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 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.
How it starts
The opening of the file, as written. The whole thing — 663 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md: SiftStack developer guide
This file guides Claude Code and developers working on SiftStack. It covers what the system is, how the repo is laid out, how to run, test and deploy each part, and the API contracts and traps that cost real time to find. The skill library's install guide is README.md; vendor contracts in depth are in docs/api/.
What SiftStack is
A real estate investing operations platform built around the DataSift.ai CRM (formerly REISift; app at app.reisift.io, API at apiv2.reisift.io). It covers the lead lifecycle end to end:
- Acquisition: county public-notice scraping (foreclosure, probate, tax sale, tax delinquent), scanned PDF and courthouse-photo import, county recorder and condemnation sources.
- Enrichment: address standardization, property data, county tax, obituary and heir research, skip trace, phone scoring.
- CRM automation: API upload into DataSift, presets and sequences, an account blueprint cloner, a sold-property guard.
- Deal analysis: comps, rehab, offer sheet, post-walkthrough package, dispo buyers, lender package.
- The agents brain: four Claude "employees" in Slack plus an autopilot that runs the prospecting lifecycle stage by stage.
- Outreach: a two-way SMS agent, Smarter Contact campaigns, cold email, direct mail.
- The skill library: Claude skills shipped as a Claude Code plugin marketplace.
Python 3.12, flat modules under src/ (the import root, not a package), three Fly.io apps with SQLite on volumes, GitHub Actions for CI and outside watchdogs.
Ground rules
- The repo is public. Never commit keys, tokens, session or cookie files, county records naming people, recordings, vendor JS bundles or seller data.
output/,logs/,.env,*_state.json,cookies.json,config/sms_*.jsonare gitignored for this reason. Readgit statusbefore every commit. - Dry by default. Anything that writes to DataSift, pays a vendor or contacts a person is a plan unless given
--commit; account-wide tools also need an env switch (*_ALLOW=1). New code keeps that contract. - One record first. Before a batch write, write one record, read it back, compare. Most DataSift and vendor failures return a success status.
- A filter is proven by a count delta. DataSift and SiftMap silently ignore unknown filter keys. Count with and without the key before trusting it; acceptance proves nothing.
- Zero is a failure. A run with zero notices, zero records or zero matches exits non-zero or says so. Never infer success from the absence of an error or of challenge markup.
- Persist "handled" only after the downstream write succeeded (seen ids, ledgers, done-files), so an aborted run retries instead of skipping work forever.
- No silent degradation. A bare
exceptthat turns a 401 or 429 into "no data" is a bug. Count errors, retry transient ones, fail the phase past a threshold. - Never fabricate. A research claim carries a URL it actually opened and a verbatim quote; a value with no source is Unknown. A wrong name or number reaches a real person.
- No em dash or en dash characters anywhere: code, docs, copy, commit messages. CI rejects them in
docs/**/*.mdandREADME.md. Number ranges use a plain hyphen. - Read large files by range. Several modules exceed 50 KB (
datasift_uploader.py,post_walkthrough.py,land_market.py,agents/autopilot/kpitracker.py,agents/autopilot/stages.py,sms_agent/selftest.py). Grep for the symbol, then read the lines around it. - A stage verified only on a workstation is not verified. After a deploy, run each changed stage once on the box: datacenter egress, missing files and missing secrets only show up there.
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.
- yesterday Changed · -186 lines · -11,937 tokens per session 6cf78b354f37
- 2mo ago First seen · 849 lines · 32,351 tokens per session scan A 2be6a80d1096
SiftStack CLAUDE.md is an instructions file published in the GitHub repository DataSift-Ty-Personal/SiftStack (0 stars, last pushed yesterday), licensed MIT. It adds 20,414 tokens to every session, about $0.0817 per session on Opus 5.5. 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 2026-08-30.
SiftStack CLAUDE.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| SiftStack CLAUDE.mdDataSift-Ty-Personal/SiftStack | Instructions file | 0 | yesterday | MIT | A | 20,414 | Claude Code |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 196k | today | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 144k | today | MIT | A | 8,130 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 196k | today | MIT | A | 5,685 | GitHub Copilot |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 150k | today | MIT | A | 4,680 | Codex, OpenCode |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 284k | today | MIT | A | 4,342 | Codex, OpenCode |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 196k | today | MIT | A | 3,153 | GitHub Copilot |