project-context-ingestion

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted content from the repository, creating a surface for indirect prompt injection. Malicious instructions embedded in build scripts, configuration files, or CI workflows could influence the agent's behavior or subsequent code generation.
  • Ingestion points: The skill reads build files (settings.gradle.kts, build.gradle.kts), configuration files (application.yml, application.properties), and CI/deployment manifests (Dockerfile, Kubernetes manifests, CI workflows).
  • Boundary markers: The skill does not define clear delimiters or instruction-ignore warnings for the ingested data.
  • Capability inventory: The skill performs extensive file reading across the repository to extract mapping data.
  • Sanitization: No sanitization or validation of the ingested repository content is mentioned.
  • [DATA_EXFILTRATION]: The skill instructions direct the agent to read files that frequently contain sensitive environment variables or configuration secrets.
  • Evidence: SKILL.md includes instructions to read .env.example, application.yml, application.properties, CI workflows, and Kubernetes manifests.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 03:29 PM
Security Audit — agent-trust-hub — project-context-ingestion