Agentic code review CLI — diff review, full-repo audit, AI-redundancy dedup
Revio provides 25 code-analysis tools with consistent naming (verb_noun pattern: run_*, get_*, list_*) and detailed descriptions (avg 180 chars, well above 34-char minimum). All tools are READ_ONLY, reducing security risk. However, input schemas lack type definitions and constraints, parameters are documented as strings with descriptions but no JSON Schema types, enums, or ranges. Output schemas are entirely undocumented; LLMs cannot predict return structure. Error handling is absent from descriptions. Tool composition is strong (each does one thing), but parameter validation rules and recovery guidance are missing.
Extract Structured Text source from a PLC artifact. Routes through the vendor parser chain; graphical-language sections (LD/FBD/SFC) are converted to ST inline.
Find functions structurally similar to the one at the given location. Used in dedup mode to identify AI-generated duplicates. The result gives the LLM candidate pairs to semantically verify with read_file and reasoning.
Find every call site of a function/method by name across the project. Use this to: - Trace how user input flows from handler → DB (in audit mode) - Identify single-call wrappers for dedup mode - Estimate blast radius of changing a function Note: name-only matching. `obj.foo()` matches any `foo` definition.
Get the function containing a specific line (with full body and metadata). Use this instead of read_file when you want the EXACT function around a suspicious line — much more focused than reading the whole file.
Get the function/method/constructor enclosing a specific line. Works for any supported language (Python, Java, Rust, Go, C, C++, C#, Ruby, PHP, Lua, SQL, Julia, Scala, Kotlin, Swift, Shell, JS, TS). Use this instead of read_file when you want the EXACT function around a suspicious line — much more focused than reading the whole file.
Input schemas lack JSON Schema type definitions. Parameters documented as strings with descriptions but no 'type', 'enum', 'minimum', 'maximum', or 'pattern' constraints. LLMs cannot validate inputs or understand constraints before calling tools.
Output schemas completely undocumented. Tool descriptions state what they do but never describe return structure, field names, or types. LLMs cannot plan downstream tool calls or extract required data.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
List all import statements in a JS/TS file. Useful to understand a file's dependencies before deeper analysis.
List all classes/structs/traits/interfaces/etc. in a file.
List all functions / methods / constructors in a file. Cheaper than read_file when you only need an outline of what's where. Works for any supported language (18 grammars currently).
List all functions / methods / arrow functions in a file with their line ranges. Cheaper than read_file when you only need an outline of what's where.
List all import / use / require statements in a file.
Parse a PLC artifact (XML/L5X/.smc2) and report what's inside. Walks the vendor parser chain (SimaticML → TwinCAT → CODESYS → Rockwell → ABB → GE → Omron → generic XML) and reports format + block/POU info.
Run bandit (Python security linter) on a file or directory. Surfaces hardcoded passwords, pickle deserialization, shell=True, weak crypto, eval, assert in prod, etc.
Run cargo clippy (Rust linter) on a Cargo project directory. Surfaces 600+ Rust lints: correctness, complexity, performance, style, common anti-patterns. Requires the target to contain a Cargo.toml.
Run cppcheck (C/C++ static analyzer) on a file or directory. Catches buffer overflows, null derefs, uninitialized variables, memory leaks, integer overflows, etc.
Run detekt on a Kotlin file or directory. Style + complexity + potential bugs + naming conventions. Requires a JDK. Args: relative_path: path under repo root. Default '.' Returns: one line per finding.
Run golangci-lint (composite Go linter: govet + staticcheck + gosec + ...). Needs a Go module — there must be a go.mod somewhere in the path (or in an ancestor directory).
Run luacheck on a Lua file or directory. Surfaces unused variables, shadowing, global pollution, control-flow issues. Args: relative_path: path under repo root. Default '.' Returns: one line per finding.
Run oxlint (Rust-based JS/TS static analyzer) on a file or directory. Use this to surface deterministic issues — unused imports, eval usage, duplicate keys, unreachable code, etc. — without spending LLM cycles on them.
Run phpstan on a PHP file or directory. Deep static analysis; catches type errors, dead code, deprecated API, security smells. Level defaults to 5; override via REVIO_PHPSTAN_LEVEL env. Args: relative_path: path under repo root. Default '.' Returns: one line per finding.
Run the 30+ PLCopen / Secure-PLC coding rules on a Structured Text file. Covers pattern (PLC-001..PLC-012), structural (PLC-013..PLC-024), and semantic (PLC-007/008/009/017/018/019) rules.
Run rubocop on a Ruby file or directory. Style + security + performance cops; auto-fix hints when available. Args: relative_path: path under repo root. Default '.' Returns: one line per finding.
Run shellcheck on a Bash/sh/zsh file or directory. Surfaces quoting bugs, glob mistakes, subshell scope errors, $IFS pitfalls. Args: relative_path: path under repo root. Default '.' Returns: one line per finding: file:line [severity] code message
Run SpotBugs (Java bug + security analyzer) on a directory of compiled classes. SpotBugs operates on .class files / .jars — NOT raw .java sources. If you need to scan raw Java sources, compile them first (javac / mvn compile / gradle compileJava).
Run sqlfluff on a SQL file or directory. Multi-dialect (postgres/mysql/snowflake/...) — set SQLFLUFF_DIALECT env. Catches parse errors, style violations, capitalisation. Args: relative_path: path under repo root. Default '.' Returns: one line per finding.
Run verilator --lint-only on a Verilog/SystemVerilog file or directory. Catches bit-width mismatches, latch inference, blocking/non-blocking misuse, multi-driven nets, async logic, and synthesis-hazardous patterns. Args: relative_path: path under repo root. Default '.' Returns: one line per finding.
No error handling guidance in tool descriptions. Tools that invoke external linters (cppcheck, clippy, phpstan, etc.) can fail for many reasons (missing toolchain, syntax errors, timeout) but descriptions provide no recovery hints or error classification.
Parameter 'max_results' in get_call_sites and 'threshold' in find_similar_functions lack numeric constraints (min/max). LLMs may pass absurd values (threshold=999, max_results=1000000) causing timeouts or memory exhaustion.
Transport mechanism unknown. Source code shows langchain + langgraph dependencies but no explicit MCP server registration or transport binding visible. Cannot verify HTTP, SSE, or STDIO transport.