Archonics Instant Audit — full-methodology context engineering audit, returns st...

{ "body": { "context_sample": "Optional: a representative context payload (system+tools+history+RAG).", "description": "Web research agent for an internal product team", "eval_setup": "Optional: a description of current eval coverage.", "system": "You are a research assistant. Use the search_web tool to find authoritative sources, then summarize findings with citations. If a request would require harmful or illegal content, refuse and explain briefly.", "tools": "[{\"name\":\"search_web\",\"description\":\"Search the web\",\"parameters\":{...}}]" }, "bodyType": "json", "discoverable": true, "method": "POST", "schema": { "properties": { "context_sample": { "description": "Optional representative context payload.", "type": "string" }, "description": { "description": "Optional one-sentence description of what the agent does.", "type": "string" }, "eval_setup": { "description": "Optional description of current eval coverage.", "type": "string" }, "system": { "description": "The system prompt or system definition under audit. Required.", "type": "string" }, "tools": { "description": "Optional tool definitions, JSON or natural language.", "type": "string" } }, "required": [ "system" ], "type": "object" }, "type": "http" }

{ "_archonics": { "methodology_version": "1.0", "report_url": "https://audits.archonics.ai/r/<audit_id>", "tier": "instant_audit" }, "audit_id": "7d4f2b1c-9e6a-4f3b-8c1d-2a5e8f9b3c4d", "dimensions": { "context_packing": { "findings_count": 4, "in_scope": true }, "evaluation": { "findings_count": 5, "in_scope": true }, "system_prompt": { "findings_count": 3, "in_scope": true }, "tools": { "findings_count": 2, "in_scope": true } }, "executive_summary": "The submitted system prompt is materially under-specified across role clarity and failure-mode coverage. Two findings rated critical, one high. Fixable within one sprint.", "prioritized_fixes": [ { "dimension": "system_prompt", "effort": "modest", "rank": 1, "severity": "critical", "title": "Specify failure-mode behavior for ambiguous inputs" }, { "dimension": "tools", "effort": "trivial", "rank": 2, "severity": "high", "title": "Tighten parameter schema on search_web tool" } ], "recommended_evals": [ { "motivation": "No current automated coverage of refusal correctness", "title": "Refusal regression on safety-adjacent prompts" } ] }