A campus lost and found management system that allows users to report lost items, report found items, view all lost/found items, and get summaries of current records.
Campus Lost & Found has 4 tools with minimal descriptions and no input validation. All tools are directly visible in main.py with @mcp.tool() decorators. Tool names follow verb_noun conventions (report_lost, report_found, show_lost, show_found), which is correct. However, descriptions are extremely brief (10-35 chars), parameter descriptions are missing entirely, and there is no output schema documentation. The schema field for each tool shows only a single 'item' string parameter with a 1-line description. No error handling guidance, no pagination, and no result limits are implemented. The server declares resources and prompts but focuses entirely on file I/O without any recovery patterns or structured output.
Log a found item to found.txt.
Log a lost item to lost.txt.
Return all reported found items.
Return all reported lost items.
All tool descriptions are under 20 characters and provide no context for when or why to use them. 'Log a lost item to lost.txt' and 'Return all reported lost items' are procedural but not instructive for an LLM deciding when to invoke the tool.
Input parameter 'item' in report_lost and report_found has only a 1-line description ('The lost item to report'). No guidance on format, length limits, or what constitutes a valid item description. LLMs may submit vague, incomplete, or nonsensical items.
show_lost and show_found return raw string output with no documented schema. LLMs cannot plan downstream processing or extract structured data (e.g., parsing item list, extracting counts). Output should be a structured object with fields like 'items: []', 'count: int', 'status: string'.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 36 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 32 | - | v1 |
No error handling guidance. If lost.txt or found.txt becomes corrupted or unreadable, the server returns a Python exception. LLM receives no actionable error message, no suggested recovery step, and no information about whether to retry.
write tools (report_lost, report_found) have no idempotency guarantee or confirmation step. If an agent retries report_lost('keys') due to a transient network error, the item is logged twice. No dry-run or confirmation pattern to prevent duplicate submissions.
show_lost and show_found return all items as a single concatenated string with no pagination. If 500 items are logged, the response will be enormous, wasting context tokens and degrading LLM reasoning. No limit enforcement and no next_cursor or offset support.