Server has 25 tools with explicit schemas and descriptions, but quality is inconsistent. All tools have input schemas (Zod-based, converted to JSON Schema) and descriptions present. However, descriptions are often terse (10-50 chars), parameter descriptions lack detail on constraints/formats, and output schemas are entirely undocumented. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Error handling delegates to subprocess exit codes with minimal recovery guidance. Naming follows verb_noun pattern consistently (feedclaw_*, dustclaw_*, etc.), which is good, but parameter naming lacks type suffixes (e.g., 'target' instead of 'feed_id_or_url'). No pagination support visible despite tools that could return large result sets (wirewatch_list, feedclaw_fetch). Composition is reasonable, tools are single-purpose, but output chaining IDs are not documented.
Analyze dependencies: outdated, unused, heavy, and duplicates.
Quick health report for a project: lines of code, file counts, language breakdown.
Show project health over time from saved snapshots.
Check a single service version across all environments.
Show only services with version drift across environments.
Full version drift report across all services and environments.
Quick overview of disk usage, free space, and top directories.
Output schemas completely undocumented. No tool documents what fields it returns, forcing LLMs to guess structure and plan downstream calls blindly.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications (READ_ONLY, WRITE, DESTRUCTIVE). LLMs cannot infer safety properties.
Parameter descriptions lack constraint details. E.g., 'Feed URL to subscribe to' does not specify format, length, or validation rules. 'Time window: 24h, 7d, 2w' should be an enum, not free text.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Deep scan with ranked list of largest files and folders.
Find known dev/OS space wasters like caches, build artifacts, and Trash.
Subscribe to an RSS/Atom feed by URL.
Generate an AI-powered digest of recent articles.
Fetch new articles from subscribed feeds.
Set up feedclaw with curated default feeds.
List subscribed RSS/Atom feeds. Optionally filter by category.
Export subscribed feeds as OPML.
Import feeds from an OPML file.
Unsubscribe from a feed by ID or URL.
List past AI analyses with risk level, model, connection count, and summary.
Send recent connections to AI for anomaly detection and risk assessment.
Show wirewatch database statistics.
List captured network connections from the database.
Show full details for a single connection by ID.
Start the wirewatch background capture daemon.
Show daemon running status, uptime, and connection count.
Stop the wirewatch background capture daemon.
No pagination support documented. wirewatch_list accepts 'limit' but no cursor/offset or total_count return. Large result sets will blow context window.
Error handling delegates to subprocess exit codes. runTool() returns stderr or generic 'exited with code N', no recovery guidance, no categorization (retryable vs fatal).