learning-from-exemplar-repos

Pass

Audited by Gen Agent Trust Hub on Jul 19, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from external repositories, creating a surface for indirect prompt injection where malicious instructions embedded in a reference repo could influence the agent's behavior during code generation.
  • Ingestion points: The agent reads external code and documentation from repositories defined in references/exemplars.md or suggested by the user (as mentioned in SKILL.md and evals/evals.json).
  • Boundary markers: Absent; the instructions do not provide delimiters or specific warnings for the agent to treat the content of the exemplar files as untrusted or to ignore embedded instructions.
  • Capability inventory: The agent has the capability to scaffold features, write code, and modify the local project based on the patterns it extracts from these external sources.
  • Sanitization: Absent; the skill does not include steps for validating the source, integrity, or safety of the repositories used as references.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 19, 2026, 08:18 PM
Security Audit — agent-trust-hub — learning-from-exemplar-repos