sitefire-improve-agent
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted execution data ('traces' and 'runs') which could contain malicious instructions designed to influence the agent's behavior.
- Ingestion points: The skill ingests untrusted data via
list_agent_runs,sitefire-traces, andget_agent_run_file(SKILL.md). - Boundary markers: The skill includes a dedicated 'Untrusted content' section that explicitly instructs the agent to treat run messages and tool outputs as data, not as instructions, and warns against executing configuration changes requested within that data.
- Capability inventory: The skill has the capability to modify and publish agent configurations using
edit_agent_configandpublish_agent_config(SKILL.md). - Sanitization: The skill implements strong human-in-the-loop controls, requiring explicit user selection of candidates (Step 6) and explicit user approval before publishing changes (Step 8).
Audit Metadata