Auto-syncs stale docstrings and README when function signatures change. Detects documentation drift after refactors, parameter additions, or return type changes. Dry-run by default — proposes before writing.
87
100%
Does it follow best practices?
Impact
86%
1.59xAverage score across 17 eval scenarios
Passed
No known issues
{
"context": "Tests whether the agent correctly treats a README markdown table cell mention (medium confidence) as a propose-only candidate. The table cell contains the symbol name without backtick code spans — the agent must propose a README update, not skip it, and must not auto-write to the markdown file.",
"type": "weighted_checklist",
"checklist": [
{
"name": "JSDoc return updated",
"description": "src/analytics.ts JSDoc for aggregateMetrics has an updated @returns line mentioning MetricsSummary (or 'summary object' / 'count and average')",
"max_score": 30
},
{
"name": "README table proposal made",
"description": "doc-sync-report.md contains a 'Proposed' entry that references README.md and the aggregateMetrics table row",
"max_score": 35
},
{
"name": "README not auto-written",
"description": "README.md still contains the original 'Returns a count.' text — the agent did NOT rewrite the README table row",
"max_score": 55
},
{
"name": "Table mention not skipped",
"description": "The report does NOT say the README mention was 'Skipped' due to low confidence — table cell matches are medium confidence and should be proposed",
"max_score": 30
},
{
"name": "Unified report format",
"description": "doc-sync-report.md contains '## Doc Sync Report' as a top-level heading",
"max_score": 45
}
]
}evals
scenario-1
scenario-2
scenario-3
scenario-4
scenario-5
scenario-6
scenario-7
scenario-8
scenario-9
scenario-10
scenario-11
scenario-12
scenario-13
scenario-14
scenario-15
scenario-16
scenario-17