Review the project for unextracted learnings and forgetting opportunities. Call this periodically (e.g., after closing several cases) to: - Find clusters of unpromoted pressure events that could become foundations - Identify cases blocking forgetting (regret 0 but unpromoted PEs remain — promote ...
AI agents call suggest_review to retrieve information from Decision OS MCP without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
suggest_review is a query and analysis tool that examines existing data (pressure events, cases, their metadata) to provide insights and recommendations. It retrieves and correlates information to identify patterns and opportunities, but does not create, modify, delete, or execute any actions. The tool is explicitly framed as a retrospective mechanism for finding unextracted learnings—a diagnostic function.
From the tool's definition Tool performs retrospective review and analysis: 'Find clusters of unpromoted pressure events', 'Identify cases blocking forgetting', 'Flag high-regret cases'.
Attacks that exploit this kind of access
Review the project for unextracted learnings and forgetting opportunities. Call this periodically (e.g., after closing several cases) to: - Find clusters of unpromoted pressure events that could become foundations - Identify cases blocking forgetting (regret 0 but unpromoted PEs remain — promote or discard to unblock) - Flag high-regret cases with no PEs (possible missed captures) This is the retrospective mechanism. Knowledge lives in foundations, not cases. It is categorised as a Read tool in the Decision OS MCP MCP Server, which means it retrieves data without modifying state.
Register the Decision OS MCP server in PolicyLayer and add a rule for suggest_review: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Decision OS MCP. Nothing to install.
suggest_review is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the suggest_review rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for suggest_review. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
suggest_review is provided by the Decision OS MCP server (marianstefi20/decision-os-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →