{"id":"799a618a-765f-4924-9935-f0fa0a700af9","arxiv_id":"2603.20380","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":4.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors propose a portable CAT data layer for managing multi-agent harnesses via plain-text files and evaluate its use with 22 models on 115 practical tasks using the npcsh shell.","lead":"The paper introduces a context-agent-tool (CAT) data layer of interrelated plain-text files to manage agent tools and behaviors in multi-agent systems, inspired by the ALARA radiation safety principle, and demonstrates it via a new command-line shell called npcsh across 22 models and 115 tasks. A smart generalist might read it to see a practical approach for making complex AI agent teams more controllable, shareable, and maintainable without relying on fragmented instructions","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Evaluation shows model performance on tasks using npcsh but supplies no comparisons or metrics testing CAT's claimed scalability, versionability, or outperformance over prose/configs.","rationale":"Reader's weakest assumption matches the gap exactly; the full-text evaluation section tests downstream model behavior rather than the harness properties that motivate the CAT layer, so the concern is load-bearing for the central claim.","tokens_in":1826,"tokens_out":308,"duration_ms":25355,"concrete_test":"Re-run the 115 tasks once with the CAT file declarations and once with equivalent prose instructions embedded in the system prompt (same models, same npcsh harness); compare success rates, error categories, and any qualitative notes on prompt length or edit effort. A difference <10% or prose performing comparably would indicate the outperformance claim lacks empirical support.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's core argument is that interrelated plain-text CAT files apply ALARA to context to deliver a scalable, versionable, collaboratively maintainable tool-access mechanism superior to fragmented prose instructions or framework-internal configs. The reported evidence consists of 22 models run on 115 tasks (file ops, web search, scripting, tool chaining, delegation) for ~2500 executions, plus characterization of model-family breakdowns. No ablation, baseline runs with prose prompts, version-control usage examples, collaboration metrics, or maintenance-effort measurements appear; thus the practical advantages asserted for the data layer itself remain unexamined.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes a Context-Agent-Tool (CAT) data layer of interrelated plain-text files that applies the ALARA principle to context in multi-agent systems. This is presented as a scalable, versionable, and collaboratively maintainable mechanism for declaring and modifying agent tool access, superior to fragmented prose instructions or framework-internal configurations. The authors implement the approach in the npcsh command-line shell and report an evaluation of 22 locally-hosted models (0.6B–35B parameters) on 115 tasks spanning file operations, web search, multi-step scripting, tool chaining, and multi-agent delegation, across approximately 2500 total executions, with characterization of model-family performance.","tokens_in":1981,"tokens_out":493,"duration_ms":49525,"significance":"If the claimed advantages of the CAT data layer hold, the work could meaningfully advance practical engineering of portable, composable multi-agent teams by aligning harness management with standard software practices such as version control and collaborative editing. The broad empirical runs across model scales and task categories provide a useful snapshot of local-model agentic capabilities, which is a concrete strength for an engineering-focused contribution.","major_comments":[{"comment":"Evaluation (as described in the abstract and results): The reported ~2500 executions demonstrate task completion with npcsh but include no baseline comparisons against prose instruction files or existing framework configurations, nor any metrics, ablations, or examples quantifying scalability, versionability, collaborative maintenance effort, or outperformance. These omissions are load-bearing for the central claim that the CAT layer delivers superior practical advantages.","section":"Evaluation"},{"comment":"Experimental design: No details are supplied on controls, success/failure criteria, error bars, statistical analysis, or how the 115 tasks were selected and scored. This prevents assessment of whether the results reliably support the assertion of enabling 'real agentic usage'.","section":"Methods"}],"minor_comments":[{"comment":"Abstract: The expansion and application of 'ALARA' to context is introduced without a concise parenthetical reminder, which may reduce immediate accessibility for readers outside radiation-safety contexts.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"This appears to be a systems/engineering contribution rather than a theoretical one; verify fit with the journal's scope for multi-agent systems work."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive report and for recognizing the potential of the CAT data layer to advance practical multi-agent harness engineering. We address each major comment below, indicating the revisions we will make to improve the manuscript while preserving its engineering focus and empirical characterization of local models.","responses":[{"response":"We acknowledge that the evaluation section emphasizes demonstration of functional agentic capability rather than head-to-head comparisons. The core argument for the CAT data layer rests on its structural properties: plain-text files enable direct use of version control, diffing, and collaborative editing, which prose instructions and opaque framework configs do not. The ~2500 executions across 22 models and 115 tasks serve to show that this layer supports real usage at scale. To strengthen the presentation, we will add a dedicated discussion subsection with concrete examples of versionability (e.g., git-tracked changes to tool-access declarations) and collaborative maintenance, plus a small-scale quantitative illustration on a subset of tasks measuring file-modularity and edit effort relative to equivalent prose prompts. A full comparative ablation across all tasks lies outside the current scope and will be noted as future work.","revision_made":"partial","referee_comment":"[Evaluation] Evaluation (as described in the abstract and results): The reported ~2500 executions demonstrate task completion with npcsh but include no baseline comparisons against prose instruction files or existing framework configurations, nor any metrics, ablations, or examples quantifying scalability, versionability, collaborative maintenance effort, or outperformance. These omissions are load-bearing for the central claim that the CAT layer delivers superior practical advantages."},{"response":"We agree that the current manuscript under-specifies the experimental protocol. In revision we will expand the Evaluation section to describe: task curation (115 tasks drawn from representative practical categories: file operations, web search, multi-step scripting, tool chaining, and delegation); explicit success criteria per category (e.g., correct file state for I/O tasks, factual accuracy for search, successful delegation for multi-agent cases); scoring procedure (automated log checks supplemented by author verification on a stratified sample); and controls (identical system prompts, tool schemas, and environment across all models). Because the study is observational—intended to characterize model-family behavior rather than test hypotheses—we did not compute error bars or perform statistical tests; we will state this explicitly and report results as descriptive success rates by category and model scale. These additions will make the support for 'real agentic usage' more transparent.","revision_made":"yes","referee_comment":"[Methods] Experimental design: No details are supplied on controls, success/failure criteria, error bars, statistical analysis, or how the 115 tasks were selected and scored. This prevents assessment of whether the results reliably support the assertion of enabling 'real agentic usage'."}],"tokens_in":1453,"tokens_out":597,"duration_ms":77179,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main thing here is a plain-text CAT file format that lets teams declare which tools each agent can use, plus the npcsh shell that loads those files and runs the agents. They map the ALARA idea to context limits and show the setup working across 22 models from 0.6B to 35B on 115 tasks like file ops, web search, and tool chaining, with roughly 2500 total runs. That produces some model-family breakdowns that could be useful for practitioners picking local models for agent work.","headline":"The paper gives a workable file-based layer for declaring agent tools in multi-agent teams and runs it on many local models, but the tests do not check whether the layer actually improves scalability or maintenance over plain prompts.","tokens_in":2517,"tokens_out":194,"would_cite":false,"duration_ms":18708,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"grok-4.3","evidence":[{"relation":"unclear","rs_module":"IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean","rs_theorem":"reality_from_one_distinction","paper_passage":"Applying the ALARA principle from radiation safety ... we introduce a declarative context-agent-tool (CAT) data layer expressed through interrelated files that scope each agent’s tool access and context to the minimum its role requires"},{"relation":"unclear","rs_module":"IndisputableMonolith/Cost/FunctionalEquation.lean","rs_theorem":"washburn_uniqueness_aczel","paper_passage":"The Jinx list enforces its constraint structurally rather than interpretively. ... tools not on the Jinx list do not exist in the agent’s schema"}],"headline":"Declarative CAT/ALARA agent harness unrelated to RS distinction-forcing or J-cost machinery","alignment":"orthogonal","rationale":"The paper's central construction is a filesystem-based declarative data layer (context/NPC/Jinx files) that structurally enforces least-privilege tool scoping via ALARA/least-privilege, evaluated on 115 tasks across 22 models. This has no connection to the RS forcing chain from a single distinction (∃x y, x≠y) through J(x)=½(x+x⁻¹)−1, φ-ladder, 8-tick periodicity, or parameter-free derivation of c/ℏ/G. No ratio-symmetric cost, golden-ratio identities, or recognition-cost reasoning appears; the work is standard multi-agent engineering.","tokens_in":48935,"confidence":"high","tokens_out":351,"duration_ms":20448,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Interrelated plain-text files apply the ALARA principle to context so users can directly declare and modify tool access for each agent in multi-agent teams.","keywords":["multi-agent systems","agent harness engineering","ALARA principle","context engineering","tool access declaration","plain-text configurations","composable agent teams","local model evaluation"],"falsifier":"A side-by-side test in which two teams maintain the same multi-agent project for several weeks, one using CAT plain-text files and the other using prose instructions, then measure the number of unintended tool calls and the time required to propagate a change across all agents.","tokens_in":2703,"feed_emoji":"🤖","tokens_out":880,"duration_ms":47676,"temperature":0.7,"pith_summary":"The paper introduces a context-agent-tool data layer built from interrelated plain-text files to manage what agents can do in multi-agent systems. It draws on the ALARA principle of keeping exposures as low as reasonably achievable, this time applied to context, so that tool access declarations become explicit, shareable, and easy to change without relying on scattered prose instructions or hidden framework settings. A sympathetic reader would care because current fragmentation makes it hard for individuals or teams to coordinate agent behaviors reliably or to maintain them as projects evolve. The authors show the approach works in practice through a command-line shell called npcsh that loads teams and runs agents, then test it with 22 local models on 115 tasks covering file work, searches, scripting, and delegation. The evaluations also map where different model sizes succeed or fail across thousands of runs.","feed_headline":"Plain-text CAT files apply ALARA to make agent tool access directly editable","feed_subtitle":"Interrelated files replace fragmented instructions so teams can version and maintain what each agent is allowed to do, shown on 22 models.","key_machinery":"The CAT data layer of interrelated plain-text files that declare context, agents, and tools for each team member, which carries the argument by turning fragmented behavioral specifications into a single, versionable, and directly editable structure.","core_discovery":"Applying the ALARA principle from radiation safety to context, we introduce a context-agent-tool (CAT) data layer expressed through interrelated plain-text files, allowing users to directly declare tool access for each agent and to modify the tools themselves that are used by the agents when processing. We demonstrate capability of this CAT data layer to enable real agentic usage by using a command-line shell that loads the team and executes agent runs -- npcsh -- and evaluating 22 locally-hosted models from 0.6B to 35B parameters across 115 practical tasks spanning file operations, web search, multi-step scripting, tool chaining, and multi-agent delegation. We characterize which model fami","pith_inferences":["The plain-text approach would integrate naturally with existing version-control workflows, letting agent configurations participate in the same review and branching processes as code.","Extending the same ALARA-style minimalism to other context elements such as memory stores or output formats could further reduce unintended agent actions.","The evaluation method of running thousands of executions across model sizes offers a template for future benchmarks that test harness engineering rather than isolated model capability.","Adoption in domains with strict audit requirements could follow directly because every tool access declaration is human-readable and traceable."],"forward_implications":["Agent behaviors become portable across projects because the harness specifications live in ordinary files rather than inside any particular framework.","Teams can coordinate context engineering work by editing and reviewing the same set of plain-text declarations.","Individual human-agent interactions improve because tool access can be adjusted precisely for each agent instead of relying on general instructions.","Model selection for practical work becomes more informed once performance patterns across task categories are measured under the same harness.","Tool chaining and delegation steps gain reliability when each agent's allowed actions are stated explicitly in the data layer."],"fun_headline_variants":["CAT files apply ALARA for editable agent tool access","Plain-text CATs replace fragmented agent instructions","ALARA applied to context via plain-text CAT data layer","CAT data layer allows direct tool declarations for agents"],"cache_read_input_tokens":64,"weakest_assumption_plain":"That plain-text files applying the ALARA principle to context will deliver a more scalable, versionable, and collaboratively maintainable way to declare and change agent tool access than prose instructions or internal framework configurations.","fun_headline_variants_meta":{"raw":{"variants":["CAT files apply ALARA for editable agent tool access","Plain-text CATs replace fragmented agent instructions","ALARA applied to context via plain-text CAT data layer","CAT data layer allows direct tool declarations for agents"]},"model":"grok-4.3","cost_usd":0.008963,"raw_usage":{"total_tokens":4002,"prompt_tokens":780,"num_sources_used":0,"completion_tokens":59,"cost_in_usd_ticks":89628000,"prompt_tokens_details":{"text_tokens":780,"audio_tokens":0,"image_tokens":0,"cached_tokens":64},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3163,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":780,"tokens_out":59,"duration_ms":64068,"temperature":1.0,"reasoning_tokens":3163,"cache_read_input_tokens":64,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-21T11:07:33.644818+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A side-by-side test in which two teams maintain the same multi-agent project for several weeks, one using CAT plain-text files and the other using prose instructions, then measure the number of unintended tool calls and the time required to propagate a change across all agents.","supporting_citations":[],"review_version":1}