replay-user-flows
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data stored in the
.reticle/flows/directory to replay user journeys. This design allows the agent to ingest and act upon data that could theoretically be modified to influence agent behavior. - Ingestion points: Flow files located in
.reticle/flows/and application signals used for verification. - Boundary markers: The instructions do not define specific delimiters or security warnings to prevent the agent from obeying instructions potentially embedded within the flow data.
- Capability inventory: The agent uses tools like
reticle_verifyandreticle_runwhich can execute various actions and sub-tools based on the content of the ingested flows. - Sanitization: No explicit sanitization or validation of the flow content is mentioned in the instruction file.\n- [EXTERNAL_DOWNLOADS]: The skill provides links to fetch technical references and documentation from the vendor's official domain.\n
- Evidence: Provides
curlcommands to accesshttps://docs.reticle.sh/flows.mdandhttps://docs.reticle.sh/llms.txt.
Audit Metadata