social-media-analyzer
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes files 'eval_result.json' and 'skill-report.json' that contain self-referential content claiming the skill is safe-to-publish and that any security findings are false positives. This content targets the behavior of security analyzers to bias their verdict (Category 8e).
- Ingestion points: Ingests social media campaign metrics (likes, shares, clicks) via JSON and CSV data as documented in SKILL.md and calculate_metrics.py.
- Boundary markers: Absent; there are no instructions to isolate or delimit potentially untrusted input data.
- Capability inventory: The skill only performs mathematical calculations for ROI and engagement rates. It lacks network access, file writing capabilities, and subprocess execution tools.
- Sanitization: Absent; metrics are used directly in arithmetic operations without validation.- [METADATA_POISONING]: The skill provides inconsistent and potentially deceptive authorship information. While the platform metadata identifies the author as aiskillstore, the internal report files attribute the skill to Community and alirezarezvani. Furthermore, the reports reference non-existent technologies (e.g., gpt-5.5), indicating fabricated metadata.
Audit Metadata