Audit npm, Yarn, and pnpm lockFiles as both an MCP server and a CLI tool.
locklens has 8 tools with basic schemas and descriptions, but significant gaps prevent production use. All tools have descriptions (10-80 chars, mostly adequate), and input schemas are present with type definitions. However, output schemas are completely undocumented, LLMs cannot plan downstream calls or extract results. Parameter descriptions lack actionable constraints (no enums, ranges, or format specs). Error handling is absent from tool definitions. Tool names are verb-based but lack clarity: 'resolveAuditSource' vs 'resolveRemoteConnectivityTarget' are confusingly similar. No tool annotations (readOnlyHint/destructiveHint) despite clear read/write distinctions. Composition is weak: 7 of 8 tools are READ_ONLY discovery/validation steps, with only 1 WRITE tool (prepareAuditWorkspace), suggesting incomplete workflow design.
Check TCP connectivity to remote repository host
Clean up expired temporary artifacts from previous audit runs
Detect lockfile type and location in a directory
Prepare audit workspace by cloning or fetching remote repository
Resolve audit source to either local directory or remote Git repository
Resolve remote repository URL to hostname and port for connectivity check
Determine which provider to use for fetching remote workspace
Run package audit on lockfiles for npm, Yarn, and pnpm
No output schemas documented for any tool. LLMs cannot infer result structure, plan chaining, or extract required fields for downstream calls.
Parameter descriptions lack actionable constraints. 'threshold' accepts 'low|moderate|high|critical' but no enum declared. 'retryCount' has no min/max bounds. 'timeoutMs' has no guidance on reasonable values.
Tool names are ambiguous and confusing. 'resolveAuditSource', 'resolveRemoteConnectivityTarget', and 'resolveRemoteWorkspaceProviderPlan' all start with 'resolve' but do different things. LLMs will struggle to distinguish them.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint/destructiveHint/idempotentHint) despite clear semantic differences. 7 tools are READ_ONLY, 1 is WRITE, this distinction should be machine-readable.
No error handling guidance in tool definitions. What happens if a lockfile is not found? If Git connectivity fails? If audit-ci crashes? LLMs have no recovery path.