shot-prompt
Pass
Audited by Gen Agent Trust Hub on Aug 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting untrusted data from external project files and incorporating it into its logic and output.
- Ingestion points: The skill reads content from multiple untrusted sources including files in
/prompts/,/docs/registry.md,/docs/breakdown.md, and asset passports located under/assets/. - Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to disregard instructions that might be embedded within the ingested shot cards or asset descriptions.
- Capability inventory: The agent is granted capabilities to write new files to
/prompts/, update logs in/docs/generation-log.md, and organize files into/selects/and/generations/directories based on the processed data. - Sanitization: The skill explicitly instructs the agent to copy canonical descriptors 'verbatim' and mandates that it should 'not shorten, paraphrase, or clean up' the content, which prevents the agent from naturally filtering out potential malicious instructions embedded in the source text.
Audit Metadata