# Oh My Sol

> Documentation for Oh My Sol (OMS), a local Solana coding harness from barestate.

## Docs

- [Oh My Sol](/docs/index): The Solana harness, batteries included. A terminal app for building on Solana with the model you already use.
- [Quickstart](/docs/quickstart): Install OMS, start it, connect a model provider and sign in. Each step says what to do if it fails.
- [Connect a model](/docs/guides/connect-a-model): Add a provider with /connect, import keys from a file, and pick a model with /model.
- [How a task runs](/docs/concepts/how-a-task-runs): What happens between your request and a closed task, and the limits that stop a run. Read it to know why OMS stops when it does.
- [The independent checker](/docs/concepts/independent-checker): Only an offline verifier can close a task. Read this to learn what it checks, what 'verified' means, and where it can still be wrong.
- [The local chain](/docs/concepts/local-chain): How OMS runs Surfpool 1.7.0 offline on your machine, and how checkpoints and rewind work. Read it when the chain misbehaves or you want to know what it can reach.
- [Project detection](/docs/concepts/project-detection): How OMS works out your project's toolchain by reading files, without running any of your code.
- [Error decoding](/docs/concepts/error-decoding): How OMS turns a custom program error or a failed transaction into a name and a message, and the order it looks in.
- [Search](/docs/concepts/search): How OMS looks things up, Solana sources first and then the web, and the limits on what it fetches.
- [Solana references](/docs/concepts/solana-reference): The program reference built into OMS and the pinned Solana documentation set it draws on, so the model starts with facts current to the pinned versions.
- [Work until a check passes](/docs/guides/work-until-it-passes): Use /goal, /verify and /status to run a task against an independent check.
- [Decode an error](/docs/guides/decode-an-error): Turn a custom program error code or a transaction signature into a name, a message and the failing instruction.
- [Use the local chain](/docs/guides/use-the-local-chain): Check health, fund an address, read a transaction, and rewind the local Surfpool chain.
- [Check your toolchain](/docs/guides/check-your-toolchain): Compare your project with the pinned Solana toolchain and apply the proposed fixes in a worktree.
- [Run the offline security review](/docs/guides/security-review): A static review of a local Cargo package. No model. Not an audit.
- [Slash commands](/docs/reference/slash-commands): All 43 OMS slash commands, grouped by what they do.
- [CLI](/docs/reference/cli): The oms command, its subcommands and flags.
- [Configuration](/docs/reference/configuration): Files, folders and environment variables OMS reads.
- [Models and providers](/docs/reference/models): The model catalog in OMS 0.1.0.
- [Security overview](/docs/security/overview): What protects what, what is wired into the shipped app, and which claims have tests.
- [Signer](/docs/security/signer): A separate signing process built so that the model never sees key bytes. Built and tested, not connected to the shipped app.
- [File-read guards](/docs/security/file-read-guards): Two layers that keep the model away from key folders and the pinned verifier, including through .. and symlinks.
- [Shell guard](/docs/security/shell-guard): A text scan over shell commands that blocks mentions of denied paths, and the four cases it is known to miss.
- [Sandbox](/docs/security/sandbox): macOS Seatbelt, Docker mode, and what Linux host mode does not have.
- [Credentials](/docs/security/credentials): How OMS stores provider keys and the account token, and what the model never sees.
- [Benchmarks overview](/docs/benchmarks/overview): The headline results, how to read them, and what they do not show.
- [Fresh tasks](/docs/benchmarks/fresh-tasks): Pass rates on eight offline tasks that need Solana facts newer than the models' training data.
- [Solana Bench](/docs/benchmarks/solana-bench): Scores for the round-16 OMS build against Codex CLI and Claude Code, how significance was tested, and the earlier series.
- [Methodology](/docs/benchmarks/methodology): How the tasks were built and selected, which versions ran, when, and how pass and fail were decided.
- [Changelog](/docs/changelog): What changed in OMS, from the first public release.