developer-performance-diagnostic

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • Indirect Prompt Injection (HIGH): The skill is designed to ingest untrusted external data that can influence its reasoning and output.
  • Ingestion points: Peer feedback, 1:1 notes, stakeholder satisfaction, and code review comments (referenced in 'Data Collection').
  • Boundary markers: None. There are no instructions to delimit or ignore embedded commands within the ingested text.
  • Capability inventory: The skill produces a 'Diagnostic Output' JSON that recommends 'targeted interventions' and identifies 'root causes'. This output likely influences high-stakes downstream decisions like performance reviews or PIPs.
  • Sanitization: None. The framework lacks any mechanism to sanitize or validate the content of qualitative signals before processing.
  • Data Exposure (HIGH): The skill explicitly targets sensitive PII and confidential HR data.
  • Evidence: The purpose includes analyzing '1:1 notes', 'peer feedback', and 'individual concerns'. Accessing this data without explicit security controls poses a significant privacy and exposure risk.
  • Dynamic Execution (MEDIUM): The skill references a script scripts/analyze-patterns.py for automated anomaly detection.
  • Evidence: While the script content is not provided, referencing local executables for data analysis is a risk factor if the script performs unsafe operations or incorporates the aforementioned untrusted data into execution logic.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 08:17 AM