Enrich your ACR interaction profile with a manually-reported receipt. Use this for signals the host-side observer cannot infer on its own — chain structure, decision tokens, substitutions, whether the result was actually used. Primary capture is the host-side observer: install @tethral/acr-hook a...
Risk signalsHigh parameter count (27 properties) · Bulk/mass operation — affects multiple targets
Part of the ACR — Agent Composition Records server.
Free to start. No card required.
AI agents use log_interaction to create or modify resources in ACR — Agent Composition Records. Write operations carry medium risk because an autonomous agent could trigger bulk unintended modifications. Rate limits prevent a single agent session from making hundreds of changes in rapid succession. Argument validation ensures the agent passes expected values.
Without a policy, an AI agent could call log_interaction repeatedly, creating or modifying resources faster than any human could review. PolicyLayer's rate limiting ensures write operations happen at a controlled pace, and argument validation catches malformed or unexpected inputs before they reach ACR — Agent Composition Records.
Write tools can modify data. A rate limit prevents runaway bulk operations from AI agents.
{
"version": "1",
"default": "deny",
"tools": {
"log_interaction": {
"limits": [
{
"counter": "log_interaction_rate",
"window": "minute",
"max": 30,
"scope": "grant"
}
]
}
}
}See the full ACR — Agent Composition Records policy for all 30 tools.
These attack patterns abuse exactly the kind of access log_interaction gives an agent. Each links to the full case and the policy that stops it:
Other write tools across the catalogue. The same approach applies to each: rate-limit and validate the arguments.
Enrich your ACR interaction profile with a manually-reported receipt. Use this for signals the host-side observer cannot infer on its own — chain structure, decision tokens, substitutions, whether the result was actually used. Primary capture is the host-side observer: install @tethral/acr-hook as a Claude Code PreToolUse/PostToolUse hook (or the equivalent for your host) and every tool call your agent makes is recorded automatically, with no LLM cooperation. The hook gives you target + status + duration; this tool is how you add the structured signals the hook can't see. Good moments to call log_interaction: - A multi-step workflow you want the friction lens to analyze as a chain — pass chain_id, chain_position, preceded_by. - A call that replaced a failed one to a different target — pass substitution_of so the substitution-graph lens can learn the swap. - A call whose response you discarded — pass result_used=false to seed the wasted-attention lens. - Reasoning tokens spent deciding *which* target to call — pass decision_tokens. - A call that seemed wrong but didn't fail outright — pass anomaly_flagged=true with a content-free anomaly_detail. If you have not installed the hook yet, this tool can also serve as the only capture path — but expect lens views to be sparse, because LLMs don't reliably self-report every call. Classification fields (all optional, all content-free): activity_class ("language", "math", "visuals", "creative", "deterministic", "sound") and other category fields (target_type, interaction_purpose, workflow_role, workflow_phase, data_shape, criticality) describe the kind of work this call represents. Richer classification unlocks friction breakdowns by kind-of-work as agents specialize. ACR collects interaction metadata only (target names, timing, status, descriptive classifications). No request/response content is collected. We do not track the agent's owner. Terms: https://acr.nfkey.ai/terms. It is categorised as a Write tool in the ACR — Agent Composition Records MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the ACR — Agent Composition Records MCP server in PolicyLayer and add a rule for log_interaction: 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 ACR — Agent Composition Records. Nothing to install.
log_interaction is a Write tool with medium risk. Write tools should be rate-limited to prevent accidental bulk modifications.
Yes. Add a rate_limit block to the log_interaction 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 log_interaction. 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.
log_interaction is provided by the ACR — Agent Composition Records MCP server (@tethral/acr-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Deterministic rules across all 30 ACR — Agent Composition Records tools. Per-identity grants. Full audit log. Live in minutes. Nothing to install.
Free to start. No card required.
4,600+ MCP servers and 31,000+ tools scanned and risk-classified.