Local-first semantic knowledge graph for code navigation and AI-assisted development
Rig Constellation exposes 36 tools for code graph navigation and manipulation. Only 6 tools have visible input schemas in the source (rig_callers, rig_deps, rig_files, rig_neighbors, rig_node, rig_search); the remaining 30 are defined in .contracts/tools/*.contract.ts files that are not provided. Of the 6 visible tools, only rig_search has a substantive description (>20 chars). Tool names follow verb_noun convention (rig_add_*, rig_list_*) but lack clarity on domain boundaries, e.g., rig_explore, rig_drill, rig_chain are vague. No visible error handling guidance, no output schema documentation, and no parameter descriptions for most tools. The contract system suggests structured definitions exist, but without access to the contract files, schemas cannot be verified. Conservative scoring reflects inability to audit 83% of tool definitions.
List functions that directly call the given function. One-hop.
30 of 36 tools lack visible descriptions and input schemas. Contract files (.contracts/tools/*.contract.ts) are not provided in source, making verification impossible. Per HARD SCORING RULES, tools without visible schemas score 0 for schema dimension.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 33 | 2025-06-18+ | v2 |
List declared dependencies. Optional manifest_path filter.
List indexed files, optionally substring-filtered.
All edges incident to an anchor in both directions. {src, dst, kind, table}.
Return metadata for a single anchor: name, kind, file_path, signature, docstring, neighbor counts.
FTS5 search over node names + aliases. Returns ranked anchors {id, name, kind, file_path}. Optional kind filter.
Declared deps with no matching imports in source. Heuristic.
Vague tool names reduce LLM clarity on intent. rig_explore, rig_drill, rig_chain, rig_cold, rig_focus lack action verbs or domain specificity. LLMs cannot infer whether these read, write, or transform state.
No visible error handling guidance in any tool definition. LLMs cannot determine if failures are retryable, user-fixable, or fatal. No recovery paths documented.
Output schemas not documented for any tool. LLMs cannot plan downstream tool calls or extract required fields (e.g., node IDs for chaining). Baseline: 100% of A+ tools document return types.
No parameter descriptions visible for 6 tools with schemas (rig_callers, rig_deps, rig_files, rig_neighbors, rig_node, rig_search). LLMs cannot infer parameter meaning from names alone. Baseline: 100% of A+ tools have param descriptions.