Skip to content
@cmdrvl

CMD+RVL

Deterministic tools for data accountability. Know what changed, prove what was known, refuse when uncertain.

CMD+RVL

Open-source CLI tools that give agents facts instead of guesses about data.

CMD+RVL builds systems for reasoning over changing data. These repos are the open-source layer: small, deterministic tools that check whether two files can even be compared, say exactly what changed, check rules, score quality, and decide proceed/escalate/block — each one backed by a declared, versioned artifact (a rule set, a registry, a schema) instead of logic buried in code.

Give this to your agent

The fastest way to understand these tools is to hand them to your coding agent. Paste SYSTEM_PROMPT.md into its context, or point it at AGENT_PROMPT.md for the full guide. Every tool also answers --describe and --schema directly, so an agent can explore on its own from there.

Use these in conjunction with our MCP servers — the tools verify and seal what your agent already has locally, the MCP servers give it public and licensed data to check against.

What changes

Without these tools, an agent guesses:

  • "I think these files are probably comparable."
  • "It looks like only a few things changed."
  • "This output seems mostly fine."

With them, it gets answers:

  • shape — are these two files actually comparable?
  • rvl — what materially changed, and by how much?
  • verify — did this pass the declared rules?
  • benchmark — how good is this against a gold set?
  • assess — proceed, escalate, or block?

Example: an agent gets two quarterly CSVs and needs to know if anything important changed. shape confirms they're comparable, rvl reports 3 of 847 rows changed with one new position, and pack can seal that as evidence to review or share later. Instead of a vague summary, you get a small, checkable trail.

Tools

Tool What it does
shape Checks whether two datasets are structurally comparable before you diff them
rvl Finds the material numeric differences that explain what changed
verify Checks declared rules and constraints over files or relations
benchmark Scores candidate outputs against a gold set
assess Turns evidence into a deterministic proceed / escalate / block call
canon Resolves entity identifiers against versioned registries
veil Keeps raw sensitive files out of an agent's context while still allowing authorized subprocess workflows
airlock Proves what did and didn't cross into a model's context
vacuum Enumerates files into a deterministic manifest
hash Adds SHA-256/BLAKE3 content identity to a manifest (installs as hashbytes)
fingerprint Matches files against versioned template definitions
profile Defines and freezes a column-scoping config for the tools above
lock Pins a dataset into a self-hashed, tamper-evident lockfile
pack Bundles lockfiles and reports into one sealed, content-addressed evidence pack

Typical pipeline: shapervl / verify / benchmarkassesspack. Add vacuumhashbyteslock first for stronger provenance on the inputs.

More repos: SEC/EDGAR tools, developer tools, deferred

SEC EDGAR & financial data

Tool What it does
cmdrvl-xew Detects enforcement-fragile XBRL patterns in SEC filings
edgar-change-interpreter Claude skill for spotting material changes in SEC filings
edgar-fabric-ingest Reference implementation for ingesting EDGAR disclosures into an event store

Developer tools

Tool What it does
regret Mines git history for high-precision regret signals — reverts, linked fixes, patch-id matches
twinning In-memory database twin that speaks the real wire protocol

Deferred

Tool What it does
compare Exhaustive raw diff capability, currently deferred

Install

brew install cmdrvl/tap/{shape,rvl,verify,benchmark,assess,veil,airlock}

Full set:

brew install cmdrvl/tap/{vacuum,hash,fingerprint,profile,lock,shape,rvl,verify,benchmark,assess,canon,pack,veil,airlock}

(cmdrvl/tap/hash provides the hashbytes binary; cmdrvl/tap/benchmark installs as cmdrvl-benchmark.)

Links

Pinned Loading

  1. edgar-change-interpreter edgar-change-interpreter Public

    Claude skill for identifying material changes, silent risks, and interpretation traps in SEC filings (10-K, 10-Q, 8-K). Built by CMD+RVL.

    TypeScript 6 1

  2. edgar-fabric-ingest edgar-fabric-ingest Public

    A TypeScript CLI and Codex-friendly reference implementation for ingesting EDGAR disclosures into an append-only CQRS context layer. Focused on auditability, replayability, and strict schema discip…

    TypeScript

  3. rvl rvl Public

    rvl reveals the smallest set of numeric changes that explain what actually changed between two datasets — or confidently tells you nothing changed.

    Rust 1

  4. cmdrvl-xew cmdrvl-xew Public

    CLI engine that detects enforcement-fragile XBRL patterns in SEC filings and produces reproducible Evidence Packs for filing teams and vendors.

    Python

Repositories

Showing 10 of 27 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…