AI-navigability code structure analysis — measures how expensive your code is for AI to edit
Adit provides 9 specialized static-analysis tools with clear, domain-specific descriptions (avg ~120 chars). All tools have input schemas with type definitions and parameter descriptions. However, output schemas are not documented in the visible code, and error handling guidance is absent. Tool names follow verb_noun convention (adit_score_*, adit_detect_*) but lack action verbs (should be 'analyze_' or 'detect_' rather than 'adit_'). No tool annotations (readOnlyHint, destructiveHint) despite all being read-only. Parameter descriptions are adequate but lack format/constraint details (e.g., 'Directory path' doesn't specify relative vs absolute, or whether it must exist).
List function/method/class names defined in multiple files that will produce noisy grep results
Show how many files import from a given file and which names are most widely consumed
Pre-edit briefing: returns structural warnings about a file BEFORE you edit it. Call this before editing any file to learn about cross-file risks (ambiguous names, blast radius, single-consumer imports) that you'd otherwise discover one grep at a time.
Detect import cycles that cause circular comprehension for AI agents
Compare code structure metrics between a git ref and HEAD, reporting regressions (metrics that got worse)
Show extern "C" boundary crossings for a file (Rust): both exported extern "C" fn definitions and extern "C" { } declarations, a signal for cross-language coupling invisible to use/mod import tracking
Output schemas not documented. Tools return analysis results but LLMs cannot plan downstream actions without knowing field names, types, and structure. Violates pattern:tool requirement.
Tool names lack action verbs. 'adit_score_repo' should be 'analyze_repo_navigability' or 'score_repo_for_ai_readability'. Current names are domain-specific but opaque to LLMs unfamiliar with 'adit' terminology.
No tool annotations despite all tools being read-only. Missing readOnlyHint declarations prevent LLMs from understanding safety guarantees. Violates current spec (2026-07-28) tool annotation pattern.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2026-07-28+ | v2 |
List single-consumer imports that should be co-located with their only consumer to reduce AI read cost
Score a single file for AI-navigability metrics (analyzes sibling files for cross-file metrics)
Score all files in a directory for AI-navigability metrics: context reads, grep ambiguity, file size, blast radius, import cycles
Parameter descriptions lack format constraints. 'path' params don't specify: relative vs absolute, must-exist requirement, or whether directories/files are required. Invites invalid input from LLMs.
No error handling guidance. Tools return analysis results but no documentation of failure modes (e.g., 'path not found', 'unsupported language', 'git ref invalid'). LLMs cannot recover from errors.