{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "Your coding assistant makes changes that still need substantial repair before merging.",
    "change": "Try the technique on held-out issues from a repository with reliable tests.",
    "baseline": "The same coding agent, repository snapshot, tools, and token budget.",
    "steps": [
      "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.",
      "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.",
      "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.",
      "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test."
    ],
    "measure": "Correct patches, regression failures, human repair time, and total cost per accepted change.",
    "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation."
  },
  "builder_review": null,
  "arxiv_id": "2609.20804",
  "title": "An Empirical Study of Harness Design for Coding Agents",
  "summary": "This replaces treating coding harnesses as monolithic systems with component-level selection: rule-based elision staged before LLM summarization, optional planning, and a bash-only or predefined-tool action space chosen per model. The cost is modest implementation effort for context management plus summarization tokens, and recoverable-elision machinery that models rarely use and that yields no accuracy gain. Failures include context-overflow crashes under tight budgets if management is omitted, degraded performance when bash-only interfaces are given to models with weak bash proficiency, and planning overhead that does not help weaker models on cost or stronger models on accuracy.",
  "rationale": "All components run within a fixed execution loop on top of third-party models, so the reader can adopt staged elision-plus-summarization and match the action space and planning to the chosen model's bash proficiency and context budget without extra infrastructure.",
  "technique": "Model- and budget-aware coding harness design",
  "family": "agentes_de_codigo",
  "recommendation": "adotar",
  "published": "2026-09-17",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2609.20804",
  "independent_implementations": 9,
  "total_implementations": 10,
  "stars": 30,
  "citations": 0,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "This replaces treating coding harnesses as monolithic systems with component-level selection: rule-based elision staged before LLM summarization, optional planning, and a bash-only or predefined-tool action space chosen per model. The cost is modest implementation effort for context management plus summarization tokens, and recoverable-elision machinery that models rarely use and that yields no accuracy gain. Failures include context-overflow crashes under tight budgets if management is omitted, degraded performance when bash-only interfaces are given to models with weak bash proficiency, and planning overhead that does not help weaker models on cost or stronger models on accuracy.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.20804",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "The paper reports qualitative findings rather than a single headline number: bash-only interfaces yield 'substantially lower cost' for bash-capable models, planning acts as a cost saver for stronger models with little accuracy change, and staged rule-based elision before LLM summarization gives the strongest efficiency among five context-management strategies across 176 matched settings on SWE-Bench Verified and Terminal-Bench 2.1.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.20804",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "exposure_map": [
    {
      "dimension": "quality",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compute",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "latency",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "operations",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compatibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "security",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "data_and_training",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "reproducibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "risks": [],
  "minimum_test": [],
  "open_questions": [],
  "independent_tests": [],
  "lab_reproductions": [],
  "report_available": false,
  "validation_tier": "unknown",
  "evidence_tier": "unknown",
  "infrastructure_basis": "unknown",
  "equations": [],
  "equations_status": "not_fetched",
  "core_kind": "",
  "equations_fetched_at": "",
  "equations_source_url": "",
  "equations_source_sha256": "",
  "exposures_reviewed_at": "",
  "exposures_source_sha256": ""
}