claude-flow
Version:
Ruflo - Enterprise AI agent orchestration for Claude Code. Deploy 60+ specialized agents in coordinated swarms with self-learning, fault-tolerant consensus, vector memory, and MCP integration
147 lines (129 loc) • 6.17 kB
JavaScript
/**
* ADR-112 — MCP tool description discoverability audit.
*
* Scans every MCPTool definition in src/mcp-tools/*.ts and checks each
* description for "use this over native when?" guidance signals:
* - /Use when/i
* - /Prefer .* over/i
* - /Pair with/i
* - /fall back/i
* - /use over native/i
*
* Emits a JSON report and a human-readable summary. Exits non-zero if
* the no-guidance count exceeds the baseline registered in
* verification/mcp-tool-baseline.json — the baseline is monotone-decreasing,
* so any regression fails CI.
*
* Usage:
* node scripts/audit-tool-descriptions.mjs [--update-baseline]
*
* --update-baseline writes the current no-guidance count to the baseline
* file. Run this AFTER landing a description-improvement PR and validating
* the fixes locally; do NOT use to mask regressions.
*/
import { readFileSync, readdirSync, writeFileSync, existsSync } from 'node:fs';
import { join } from 'node:path';
const REPO_ROOT = process.cwd();
const TOOLS_DIR = join(REPO_ROOT, 'v3/@claude-flow/cli/src/mcp-tools');
const BASELINE_FILE = join(REPO_ROOT, 'verification/mcp-tool-baseline.json');
const GUIDANCE_PATTERNS = [
/Use when/i,
/Prefer .* over/i,
/use .* over (native|the )/i,
/Pair with/i,
/fall back/i,
/native .* is (fine|wrong)/i,
];
const MIN_DESCRIPTION_LENGTH = 80;
function hasGuidance(desc) {
return GUIDANCE_PATTERNS.some(re => re.test(desc));
}
function scanFile(file) {
const src = readFileSync(file, 'utf-8');
// Match `name: '...',` followed by `description: '...'` (with escaped chars).
const re = /name:\s*'([^']+)',\s*\n\s*description:\s*'((?:[^'\\]|\\.)*)'/g;
const results = [];
let m;
while ((m = re.exec(src))) {
results.push({ name: m[1], description: m[2], hasGuidance: hasGuidance(m[2]) });
}
return results;
}
const updateBaseline = process.argv.includes('--update-baseline');
const files = readdirSync(TOOLS_DIR)
.filter(f => f.endsWith('.ts') && !f.endsWith('.test.ts'))
.map(f => join(TOOLS_DIR, f));
const allTools = files.flatMap(scanFile);
const withGuidance = allTools.filter(t => t.hasGuidance);
const withoutGuidance = allTools.filter(t => !t.hasGuidance);
// Additional static checks — every MCPTool description must:
// 1. Have "Use when …" guidance (the original check)
// 2. Be at least MIN_DESCRIPTION_LENGTH chars (catches near-empty descriptions)
// 3. Be unique across all tools (catches lazy copy-paste of a generic line)
const tooShort = allTools.filter(t => t.description.length < MIN_DESCRIPTION_LENGTH);
const descCounts = new Map();
for (const t of allTools) {
descCounts.set(t.description, (descCounts.get(t.description) ?? 0) + 1);
}
const duplicates = [...descCounts.entries()].filter(([, c]) => c > 1);
const report = {
scannedAt: new Date().toISOString(),
totalTools: allTools.length,
withGuidance: withGuidance.length,
withoutGuidance: withoutGuidance.length,
tooShort: tooShort.length,
duplicateDescriptions: duplicates.length,
offenders: withoutGuidance.map(t => ({ name: t.name, description: t.description.slice(0, 100) })),
shortOffenders: tooShort.map(t => ({ name: t.name, length: t.description.length, description: t.description })),
duplicateGroups: duplicates.slice(0, 5).map(([desc, count]) => ({ count, description: desc.slice(0, 100) })),
};
console.log(`MCP tool discoverability audit (ADR-112)`);
console.log(`========================================`);
console.log(`Total tools scanned: ${report.totalTools}`);
console.log(`With Use-when guidance: ${report.withGuidance}`);
console.log(`Without guidance: ${report.withoutGuidance}`);
console.log(`Too short (<${MIN_DESCRIPTION_LENGTH} chars): ${report.tooShort}`);
console.log(`Duplicate descriptions: ${report.duplicateDescriptions}`);
let baseline = { withoutGuidance: report.withoutGuidance, tooShort: report.tooShort, duplicateDescriptions: report.duplicateDescriptions };
if (existsSync(BASELINE_FILE)) {
baseline = { ...baseline, ...JSON.parse(readFileSync(BASELINE_FILE, 'utf-8')) };
console.log(`Baseline no-guidance: ${baseline.withoutGuidance}`);
console.log(`Baseline tooShort: ${baseline.tooShort ?? '(n/a)'}`);
console.log(`Baseline duplicates: ${baseline.duplicateDescriptions ?? '(n/a)'}`);
}
if (updateBaseline) {
writeFileSync(
BASELINE_FILE,
JSON.stringify({
withoutGuidance: report.withoutGuidance,
tooShort: report.tooShort,
duplicateDescriptions: report.duplicateDescriptions,
totalTools: report.totalTools,
updatedAt: report.scannedAt,
note: 'ADR-112 baseline — monotone decreasing for all three counts',
}, null, 2) + '\n',
);
console.log(`\nBaseline updated: noGuidance=${report.withoutGuidance}, tooShort=${report.tooShort}, duplicates=${report.duplicateDescriptions}.`);
process.exit(0);
}
let failed = false;
function check(label, actual, baselineValue, offendersFmt) {
if (baselineValue === undefined) return;
if (actual > baselineValue) {
console.error(`\nFAIL ${label}: count ${actual} exceeds baseline ${baselineValue}`);
console.error(` Regression: ${actual - baselineValue} new violation(s)`);
if (offendersFmt) console.error(offendersFmt);
failed = true;
} else if (actual < baselineValue) {
console.log(`\nProgress ${label}: ${baselineValue - actual} improvement(s) — run with --update-baseline to lock the new floor.`);
}
}
check('no-guidance', report.withoutGuidance, baseline.withoutGuidance,
' Add "Use when … is wrong because …" guidance per ADR-112 template.');
check('too-short', report.tooShort, baseline.tooShort,
` Top short offenders:\n ${report.shortOffenders.slice(0, 5).map(o => `${o.name} (${o.length} chars)`).join('\n ')}`);
check('duplicate-descriptions', report.duplicateDescriptions, baseline.duplicateDescriptions,
` Top duplicate groups:\n ${report.duplicateGroups.slice(0, 5).map(g => `[x${g.count}] ${g.description}`).join('\n ')}`);
if (failed) process.exit(1);
console.log(`\nOK — ${report.totalTools} tools, ${report.withGuidance} with guidance, all gates pass.`);