brida-reflex-completion-evidence-gate
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to fetch live status and documentation from the vendor's own infrastructure (e.g.,
/reflex/llms.txt,/reflex/agent.md, and/reflex/agent.json). These are documented as standard configuration checks against vendor resources. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and classify data from the local environment, including natural language completion claims and task evidence, which could be manipulated to influence the classification outcome.
- Ingestion points: Reads objective descriptions, completion claims, and deterministic evidence from the project environment as defined in
SKILL.mdandreferences/custom-reflex.json. - Boundary markers: Uses a structured JSON schema (
custom-reflex.json) and specific classification branches (verified_complete,incomplete,review) to constrain the model's interpretation. - Capability inventory: Capable of inspecting the environment and interacting with external services via MCP, SDK, or REST APIs as noted in the execution steps.
- Sanitization: Relies on deterministic host-side logic for parsing and policy enforcement, and explicitly warns against using the Reflex for granting permissions or handling sensitive credentials.
Audit Metadata