RTK Command Output Optimizer

The README says command and file contents are not collected by default and telemetry requires explicit opt-in. The Hermes plugin acts only on terminal commands and preserves the original command when rewriting fails. Install scripts, proxy forwarding, and initialization that writes configuration have external effects, while the supplied material does not provide a complete permission inventory or rollback details, limiting scores for least privilege, sensitive-data handling, rollback, and external effects. The private reporting channel and contributor security process are described, but publisher identity is unverified, so attribution remains limited.

The materials describe multi-platform CI, security checks, and fail-open behavior that preserves the original command; the Hermes test file covers several error paths. However, the CI/CD overview is only partly corroborated by the supplied workflow file, and no dependency lockfile or availability evidence is supplied, so consistency and dependency availability score low. Warning and fallback behavior are described concretely.

The README names use cases across coding agents, development commands, and operating systems, and identifies boundaries such as hooks applying only to Bash calls while built-in Read/Grep/Glob tools bypass them. Integration methods vary by agent and coverage is broad. The claim of 18 supported tools does not fully match the number shown in the table, and some environment compatibility details depend on external guides.

The README is well organized and includes installation, quick start, operation guidance, troubleshooting links, examples, and licensing information. Cargo metadata gives version and Rust requirements, and the security policy describes maintenance processes. Installation details, release history, and some limitations depend on external documentation. Apache metadata coexists with the README's BUSL licensing for the engine and proxy, so license terms must be checked by edition; maintainers are named, but publisher identity is unverified.

Command-output compression, failure summaries, a savings dashboard, and configurable pipelines show practical value for coding-agent workflows. The README distinguishes output reduction from bill reduction and says token counts use an approximate bytes/4 estimate. Savings percentages remain project claims without independent measurement in the supplied materials; readers must assess the costs and applicability of proxy traffic compression versus command-output filtering.

The README links feature, telemetry, installation, and licensing claims to specific documents; the Cargo manifest, license, security policy, and Hermes test source provide partial cross-checks. Key claims such as savings percentages and coverage are not independently verified in the supplied materials. The README does distinguish measurable bash-output reduction, approximate token counts, and total billing, which supports clear separation of facts and estimates.