Analyzes conversational stress and detects PUA (Pick-Up Artist) manipulation techniques in messages. Provides stateful stress tracking with emotional pet responses, counter-PUA education, and HTML report generation. Wraps pet_stress.py to provide stress analysis as MCP tools for Claude Code, OpenClaw, or any MCP-compatible AI client.
Pet Stress Test provides 5 tools with visible schemas and descriptions, but quality is inconsistent. Tool names follow verb_noun convention (analyze_, track_, get_, reset_, generate_). Descriptions are present but vary in clarity, some explain state effects (track_stress), others are vague (generate_html_report lacks detail on output format). Input schemas are visible for all tools, but parameter descriptions are minimal or missing. Output schemas are not documented. Error handling is absent, no guidance on recovery or retryable vs. fatal errors. Stateful session tracking (_tracker global) violates stateless request handling principle.
Analyze a single message for stress dimensions (keyword-based, instant). Returns 6 scores (0-10) + mood. Stateless — does not affect the tracker.
Create an ECharts visualization report of the session stress analysis. Generates a self-contained HTML file with interactive charts.
Get the current session stress state without adding new messages. Returns accumulated stress, dimensions, peak values, and pet emotional response.
Reset all accumulated state and start a fresh session. Returns confirmation and previous total stress.
Add a message to the session stress tracker. Returns accumulated stress, level, pet emotional response, and alerts. Call this after every user message for continuous monitoring.
No output schemas documented. LLMs cannot predict response structure or plan downstream calls. All 5 tools return JSON but structure is inferred, not declared.
Parameter descriptions are missing or trivial. 'message' param in analyze_message and track_stress lacks format/length guidance. 'role' enum in track_stress has no description of what each role means in context.
No error handling or recovery guidance. Tools return JSON errors (e.g. 'message is required') but do not classify as retryable/user-fixable/fatal or suggest next steps.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
generate_html_report description is vague ('Create an ECharts visualization report'). Does not explain output format, file location, or when to call it. 'output_path' parameter lacks guidance on valid paths or defaults.
Stateful session tracking via global _tracker violates stateless request handling. Each request should be self-contained; session state should be passed in request context or stored server-side with explicit session ID, not implicit globals.