Loading ToolBench
GitHub Actions intelligence MCP server - workflow timing analysis, configuration audit, and billing insights
Server has 4 well-named tools with complete input schemas and reasonable descriptions (avg 150 chars). All parameters are typed with descriptions and constraints (min/max, regex patterns). However, output schemas are not documented in the source code, only inferred from implementation. Error handling is present but generic (no recovery guidance). Tool descriptions lack WHEN-to-use context and dependency hints. No tool annotations (readOnlyHint, etc.) despite all tools being read-only. Naming follows verb_noun pattern well (list_*, analyze_*, estimate_*). Parameter naming is consistent (owner, repo, workflow_id, count). Missing: output schema documentation, error recovery guidance, tool annotations.
Fetch and analyze a workflow file for configuration issues, optimization opportunities, and best practices. Checks for missing caches, inefficient matrix strategies, and other common problems.
Estimate GitHub Actions billing for a repository based on recent workflow runs. Calculates billable minutes, concurrent job costs, and provides cost optimization recommendations.
Fetch timing statistics for a specific job across recent workflow runs. Shows job duration trends, step-level breakdown, and identifies performance regressions.
Fetch the last N workflow runs and compute job-level timing statistics (avg, min, max, p95) across those runs. Useful for identifying slow jobs and trends.
Output schemas not documented. Tool descriptions state what is returned (e.g., 'job-level timing statistics') but do not specify the JSON structure, field names, or types. LLMs cannot plan downstream tool calls or extract data without knowing response shape.
Tool descriptions lack WHEN-to-use context and dependency hints. E.g., 'list_workflow_performance' does not explain when to call it vs 'list_job_performance', or that it should be called first to discover available jobs. Descriptions are WHAT-focused, not guidance-focused.
No tool annotations despite all 4 tools being read-only. Missing readOnlyHint annotation would help agents understand these tools have no side effects and can be called freely without confirmation.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | 2026-07-28+ | v2 |
Error handling is generic. GitHub API errors (429, 5xx, 4xx) are caught and re-thrown with minimal context. No recovery guidance (e.g., 'rate limited, try again in X seconds' or 'workflow not found, check owner/repo/workflow_id'). LLMs cannot self-correct.
Parameter descriptions do not include format/constraint hints in text. E.g., 'workflow_id' description says 'Workflow file name (e.g. ci.yml) or numeric workflow ID' but does not state the regex pattern is enforced or what happens if both formats are passed. Constraints are in schema but not in description text.