fountain-reports
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates by filling markdown component templates located in
assets/components/with data provided by a caller or retrieved from project preferences. This architecture introduces a surface for indirect prompt injection if the input data contains malicious instructions designed to influence the agent's behavior during report generation or delivery. - Ingestion points: The skill takes "data that the preset's components need" from a caller and reads "report customizations from the Reporting section of the preferences" as described in
SKILL.md. - Boundary markers: No specific delimiters, escaping, or "ignore embedded instructions" warnings are defined in the templates to differentiate between the report structure and the interpolated external data.
- Capability inventory: The skill possesses the ability to deliver content via email using the Project API and to print content directly into the chat session.
SKILL.mdnotes that the Project API renders the markdown, which could trigger unintended actions if the markdown contains instructions the LLM is inclined to follow. - Sanitization: The instructions specifically state the skill "MUST NOT recompute, round away, or soften" the measured data, emphasizing data integrity over security filtering, and there are no instructions provided for sanitizing or validating the input data for prompt injection patterns.
Audit Metadata