File system or container scanner with Grype vulnerability detection
Single tool with basic schema and minimal descriptions. The grype-scan tool has a functional input schema (scanType enum, scanTarget string) but lacks critical metadata for production use. Description is too brief (7 words). No parameter-level descriptions visible in the tool registration, only inline Zod descriptions which are present but minimal. Output is unstructured JSON text, not a documented schema. No error handling guidance, no recovery paths, no validation messages. The tool does scan containers and filesystems, which is clear from the name, but lacks context about when to use it, what to expect, or how to handle failures. Baseline: average tool description is 194 chars; this one is 47 chars. Baseline: 100% of A+ tools have param descriptions; present here via Zod but minimal. The implementation works but does not meet production-grade quality standards for agent tooling.
File system or container scanner with Grype
Tool description is only 47 characters ('File system or container scanner with Grype'). Baseline is 194 chars (p10=34, p90=392). Missing WHAT it does (vulnerability scanning), WHEN to use it (before deployment, during security audits), and WHAT to expect (list of CVEs, severity ratings, remediation paths). LLMs cannot reliably select this tool against alternatives without richer context.
No output schema documented. The scanner returns JSON text via GrypeScan() but the response structure is never specified. LLMs cannot plan downstream usage (e.g. 'extract the vulnerability ID to file a ticket') without knowing what fields are returned. The actual return includes id, severity, description, fixAvailable, filePath but this is buried in implementation code, not surfaced as tool documentation.
No error handling guidance. The scanner calls spawnSync() and directly parses stdout as JSON. If grype is not installed, returns 'command not found' as a raw error. If the target is invalid, stderr is ignored. If JSON parsing fails, an uncaught exception propagates. LLMs receive no actionable recovery hints (e.g. 'Ensure grype is installed' or 'Verify target is a valid image or path').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | F | 42 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 42 | - | v1 |
Parameter descriptions are minimal. scanType has 'Type of scan to perform' (4 words) and scanTarget has 'Target project name or container image to scan' (8 words). Baseline is 72 chars per parameter. Missing guidance on format expectations (e.g. 'scanTarget: full image name with registry (e.g. docker.io/myorg/myapp:latest) or absolute filesystem path') and constraints (e.g. 'must be accessible to grype CLI').
No validation of scanType parameter. Although Zod enum is defined (container, file-system), if an LLM somehow passes an invalid value, there is no guard in GrypeScan() and spawnSync will interpret it as a grype flag, producing cryptic errors. Add validation and return: 'Invalid scanType: got "xyz", must be one of: container, file-system'.
Tool does not state what permissions or preconditions it requires. No mention that grype CLI must be installed, that scanTarget must be accessible, or what scan duration to expect. Agents cannot reason about whether this tool is viable in their environment.