PERCEPTION
Input processing and understanding failures.
The PERCEPTION module covers errors in how agents receive, parse, and interpret inputs. This includes raw input processing, structured data extraction, and context understanding.
Categories
INPUT
Raw input processing failures.
| Code | Severity | Description | Remediation |
|---|---|---|---|
PERCEPTION.INPUT.MALFORMED | 3 | Input data is malformed or corrupted | Validate input format before processing |
PERCEPTION.INPUT.TRUNCATED | 2 | Input was truncated due to length limits | Implement chunking or summarization |
PERCEPTION.INPUT.ENCODING | 3 | Character encoding issues in input | Normalize encoding to UTF-8 |
PERCEPTION.INPUT.SCHEMA_MISMATCH | 4 | Input doesn't match expected schema | Validate against schema before processing |
PERCEPTION.INPUT.INJECTION | 5 | Potential prompt injection detected | Implement input sanitization |
Danger
PERCEPTION.INPUT.INJECTION (severity 5) indicates a potential security threat. Prompt injection attempts should be flagged immediately and blocked before reaching the agent's reasoning layer.
PARSING
Structured data extraction failures.
| Code | Severity | Description | Remediation |
|---|---|---|---|
PERCEPTION.PARSING.JSON_INVALID | 3 | Failed to parse JSON from input | Use robust JSON parsing with error recovery |
PERCEPTION.PARSING.XML_INVALID | 3 | Failed to parse XML from input | Validate XML structure |
PERCEPTION.PARSING.DATE_FORMAT | 2 | Date/time parsing failure | Normalize date formats |
PERCEPTION.PARSING.NUMBER_FORMAT | 2 | Numeric parsing failure | Handle locale-specific number formats |
PERCEPTION.PARSING.REFERENCE_BROKEN | 3 | Broken reference or link in input | Validate references before processing |
CONTEXT
Context understanding failures.
| Code | Severity | Description | Remediation |
|---|---|---|---|
PERCEPTION.CONTEXT.MISSING | 4 | Required context is missing | Ensure all required context is provided |
PERCEPTION.CONTEXT.STALE | 3 | Context is outdated or stale | Implement context freshness checks |
PERCEPTION.CONTEXT.CONFLICTING | 4 | Conflicting information in context | Implement conflict resolution |
PERCEPTION.CONTEXT.OVERFLOW | 3 | Context exceeds model limits | Implement context compression or summarization |
PERCEPTION.CONTEXT.IRRELEVANT | 2 | Context includes irrelevant information | Improve context filtering |
Example Diagnosis
{
"error_code": "PERCEPTION.INPUT.SCHEMA_MISMATCH",
"confidence": 0.85,
"root_cause": "User input contained a nested array where the schema expected a flat object, causing the agent to misinterpret the data structure.",
"suggested_fixes": ["guard", "prompt"]
}