Loading ToolBench

Reading the current corpus…

sivru MCP server · F · ToolBench

sivru

The comprehension layer for AI-written code: authored-intent blocks, a repo-wide explainer model, and an architectural drift gate. Local, MCP-native.

2026-07-28+STDIO9 tools

47/100

Definition Quality52

Protocol Readiness35

Supportability45

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

Sivru MCP server has 9 tools with complete input schemas and descriptions, but quality is uneven. Tool names follow verb_noun convention (sivru_search, sivru_checkup, sivru_block_extract, sivru_block_edit, sivru_block_autofix, sivru_feedback_acknowledge, sivru_feedback_false_positive, sivru_feedback_suggest, sivru_explain). Descriptions are present but brief (avg ~100 chars, below the 194-char baseline). Parameter descriptions exist but lack detail on constraints, formats, and valid ranges. No output schemas documented. Error handling guidance absent. Write operations (sivru_block_edit, sivru_block_autofix, sivru_feedback_*) lack confirmation/dry-run patterns. No tool annotations (readOnlyHint/destructiveHint) visible in source.

Tools (9)

sivru_block_autofixwrite50/100

Apply the YAML-trap autofixer (E237/E238) to one file

sivru_block_editwrite50/100

Edit an authored-intent block in a file (form-driven editor with conflict detection)

sivru_block_extractread only50/100

Extract authored-intent blocks from files in a repository

sivru_checkupread only50/100

Run the checkup against a repository and return a structured report of memory file audits including claude-age, dead-reference, and skill-tools-drift checks

sivru_explainread only50/100

Build and query the codebase explainer model for architectural analysis

sivru_feedback_acknowledgewrite50/100

Acknowledge a diagnostic finding to suppress it from future reports

sivru_feedback_false_positivewrite50/100

Report a diagnostic as a false positive

sivru_feedback_suggestwrite50/100

Suggest a rewrite or improvement for a diagnostic

Issues (5)

CRITICALtool ↗

No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure.

sivru_searchsivru_checkupsivru_block_extractsivru_block_editsivru_block_autofixsivru_feedback_acknowledgesivru_feedback_false_positivesivru_feedback_suggestsivru_explain

HIGHconfirmation-request ↗

Write operations (sivru_block_edit, sivru_block_autofix, sivru_feedback_*) lack confirmation/dry-run patterns. Agents cannot preview changes before committing destructive edits.

sivru_block_editsivru_block_autofixsivru_feedback_acknowledgesivru_feedback_false_positivesivru_feedback_suggest

HIGHconstrained-input ↗

Parameter descriptions lack constraint details. 'format' enum in sivru_block_extract is documented, but most params (e.g., 'path', 'query', 'repoRoot') have minimal guidance on valid ranges, formats, or examples.

Recommendations

Document output schemas for all 9 tools. Specify field names, types, and whether results are paginated. Example: sivru_search should document {results: [{file: string, score: number, snippet: string}], total: number}.
Add constraint details to parameter descriptions. For 'path' params, clarify if relative or absolute paths are accepted, and whether symlinks are followed. For 'query' in sivru_search, specify max length and supported operators.
Implement confirmation/dry-run for write operations. sivru_block_edit should support a 'dryRun' boolean param that returns the proposed changes without committing. sivru_feedback_* tools should preview the acknowledgment before persisting.
Add error handling guidance to each tool description. Example: 'If the file is not found, returns error code FILE_NOT_FOUND. Verify the path and try again. If the path is relative, ensure it is relative to the repository root.'
Add tool annotations (readOnlyHint, destructiveHint, idempotentHint) to the MCP tool registration. Mark sivru_search, sivru_checkup, sivru_block_extract, sivru_explain as readOnly. Mark sivru_block_edit, sivru_block_autofix, sivru_feedback_* as destructive.

Score history

Overall score trend

First recorded score · v2 rubric

47/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	47	2026-07-28+	v2