auto_repair_loop
Driver: scan a subtree for cook errors, cluster them, route each cluster to the right fix (calls repair_network for structural/expression/flag issues; surfaces fix_shader / fix_reactivity as prompt hand-offs the agent must execute next turn), re-check, and iterate until clean, no-progress (stalle...
This record as markdown: /tools/hybridlabor-api-aos/auto-repair-loop.md
What auto_repair_loop does on Aos
AI agents invoke auto_repair_loop to trigger actions in Aos. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call: builds kicked off, notifications sent, workflows started.
Why auto_repair_loop is rated High
This tool executes automated repair operations on a visual system (calling repair_network, potentially modifying expressions and flags), triggering side effects that depend on detected errors and cluster routing logic. While it defaults to dry-run mode, its capability to execute repairs via repair_network and iterate until convergence or exhaustion classifies it as Execute rather than Write.
From the tool's definition The tool 'scan[s] a subtree for cook errors, cluster[s] them, route[s] each cluster to the right fix' and 'calls repair_network for structural/expression/flag issues'.
Attacks that exploit this kind of access
The rule that runs auto_repair_loop safely
PolicyLayer is an MCP gateway: it sits between your AI agents and Aos, and checks every tool call against a rule you set before the call runs. Nothing changes on the server itself. For auto_repair_loop, this is the rule to start with:
auto_repair_loop stays usable, but rate-capped: a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
The button opens the PolicyLayer dashboard: create your workspace, connect Aos, apply this rule, and every auto_repair_loop call is checked against it from then on.
Questions about auto_repair_loop
Driver: scan a subtree for cook errors, cluster them, route each cluster to the right fix (calls repair_network for structural/expression/flag issues; surfaces fix_shader / fix_reactivity as prompt hand-offs the agent must execute next turn), re-check, and iterate until clean, no-progress (stalled), or max_iterations (exhausted). Dry-run by default — one planning iteration, no writes. The loop CANNOT fix shaders or dead reactivity itself; it points the agent at them via recommended_prompts. Returns {status, iterations[], errors_before, errors_after, remaining[], recommended_prompts[], warnings}. It is categorised as a Execute tool in the Aos MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Aos MCP server in PolicyLayer and add a rule for auto_repair_loop: 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 Aos. Nothing to install.
auto_repair_loop is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the auto_repair_loop 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 auto_repair_loop. 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.
auto_repair_loop is provided by the Aos MCP server (@hybridlabor-api/aos). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
More on Aos, and thousands of servers like it.
This server
Across the catalogue