{ "schema_version": 1, "builder_plan": { "scenario": "A multi-step product task fails because the agent loses track of dependencies.", "change": "Try the proposed planning step on tasks with checkable intermediate outcomes.", "baseline": "The same agent solving the same tasks without the new planning step.", "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": "Task completion, unnecessary steps, planning tokens, response time, and total cost.", "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": "2609.20474", "title": "How Do Agent Harnesses Create Value? Planning Information and Release Control in Stateful LLM Agents", "summary": "Replaces free-form or unguided agent prompting and implicit self-judgment of completion with prewritten task-specific plans and a read-only terminal verifier that checks completion before acceptance. Costs are low: plan authoring per task type and under one cent per episode for verification, with no additional model training or infrastructure. The verifier produces false rejections (17% of correct episodes withheld), plan quality degrades on tasks outside the authored templates, and the measured gains come from Retail and Airline tool-use benchmarks and may not transfer to other domains.", "rationale": "Both components are cheap prompt- and tooling-level additions that run on a single GPU or third-party APIs, and the paper shows each delivers value independently, with the standalone verifier capturing most of the false-pass reduction at a fraction of the full stack's cost.", "technique": "Fixed task plans plus read-only terminal verifier", "family": "planejamento_e_decomposicao", "recommendation": "adotar", "published": "2026-09-17", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.20474", "independent_implementations": 1, "total_implementations": 2, "stars": 7, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "Replaces free-form or unguided agent prompting and implicit self-judgment of completion with prewritten task-specific plans and a read-only terminal verifier that checks completion before acceptance. Costs are low: plan authoring per task type and under one cent per episode for verification, with no additional model training or infrastructure. The verifier produces false rejections (17% of correct episodes withheld), plan quality degrades on tasks outside the authored templates, and the measured gains come from Retail and Airline tool-use benchmarks and may not transfer to other domains.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.20474", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Fixed plans improve oracle-verified success by 7.17 percentage points over word-count-matched sham text (90% CI 1.15-13.36, gains concentrated in higher-complexity tasks); the verifier rejects 61% of oracle-invalid Retail episodes while withholding 17% of correct ones, at under one cent per episode.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.20474", "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": "" }