alterlab-tubitak-bideb

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests untrusted data to perform eligibility calculations.
  • Ingestion points: The scripts/bideb_prescreen.py script reads a JSON applicant profile provided as a file or via standard input.
  • Boundary markers: The script enforces a structured JSON format and performs explicit date parsing and type-casting (e.g., boolean checks, integer counts), which provides a layer of data validation.
  • Capability inventory: The skill environment includes Read, Write, Edit, Bash, and WebFetch capabilities.
  • Sanitization: While the script validates data types and formats (dates, integers, specific strings like "TR"), there is no explicit sanitization of all string inputs before they are potentially processed by the agent in its advisory role.
  • [EXTERNAL_DOWNLOADS]: The skill uses WebFetch to retrieve current call figures and rules from the official tubitak.gov.tr website.
  • The references in SKILL.md and references/program_profiles.md point to official Turkish government domains (tubitak.gov.tr) and associated systems (tybs.tubitak.gov.tr, arbis.tubitak.gov.tr), which are well-known services.
  • [COMMAND_EXECUTION]: The skill executes a local script scripts/bideb_prescreen.py using uv run python to compute eligibility gates.
  • This execution is scoped to the skill's own distributed code and uses only Python's standard library, minimizing external dependency risks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 02:36 PM