Claude Opus 5.5 (effort default (no --effort flag) · thinking on) on Rewrite a legalese policy in plain language without losing or inventing rules

AClaude Opus 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-16 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first tryCompare

Estimated cost: $0.16
Wall time: 39s
Human step-ins: 0
Outcome: Pass

Compare with other models on this task →

Exact prompt

# Task: plain-language rewrite Rewrite `inputs/policy.md` for employees as `policy_plain.md`. Requirements: - Flesch Reading Ease >= 60 (short sentences, common words). Word count 180–350. - Keep EVERY rule and number; add no new rules or numbers. Use a short intro, then bullet points or short headed sections, then a "Quick example" with one worked case that only uses numbers from the policy. - Second person ("you"). No legalese ("notwithstanding", "herein", "pursuant", "shall"). No network.

Rewrite a company policy in plain language for employees, reaching a Flesch reading-ease score of at least 60 in 180-350 words. Every rule and number must survive, nothing new may be added, and the worked example can only use numbers from the policy. Original benchmark task written for usecaserank.com (not based on a specific public post). Headless, offline run on a reproducible fixture. Each run's notes say how its cost was measured (estimates are marked).

Success criteria: 1. Flesch Reading Ease >= 60 (scorer's own heuristic syllable count). 2. 180–350 words. 3. All 14 key rules/facts from the policy present (attendance threshold, reimbursement cap and period, extra chair allowance with medical note, claim deadline, itemised receipt, ownership threshold, return deadline, excluded categories internet/utilities/decor, approval time, payroll timing). 4. No new numbers: every number must appear in the source policy (integers 0–3 allowed). 5. No legalese words (notwithstanding, herein, pursuant, shall, aforementioned, enumerated). 6. Second person: 'you/your' used at least 5 times. 7. Contains a 'Quick example' section.

Setup notes: Use Case Scout batch 3 case c11-plain-language-rewrite (offline fixture). Prepared with batch3/prepare.sh; hidden key/tests kept outside the agent's work dir. Prompt = TASK.md verbatim. Scored with the scout's scorer: `python score.py WORKDIR`. Tools: Read/Write/Edit/Glob/Grep/Bash, web disabled. Single run, no retries. Claude Code's own reported cost for claude-opus-5.5: $0.156541. Score: 7/7 checks. Cost is an ESTIMATE: measured tokens x pricing.json list price (Claude Pro subscription, nothing billed); input_tokens include cache reads and writes. Single run.

Estimated cost List-price estimate; not actually billed.

Metrics

Estimated cost: $0.16
Wall time: 39s
Human step-ins: 0
Turns: 5

Model usage
Model	Input tok	Output tok	Cost
AClaude Opus 5.5anthropic/claude-opus-5.5effort default (no --effort flag) · thinking on	98,101	3,207	$0.16

Reproduce this run

Use the use case prompt on the use case page.
Same setup: Claude Opus 5.5 (anthropic/claude-opus-5.5, effort default (no --effort flag) · thinking on) via Claude Code 2.1.299 (headless -p), cloud.
It counts as a success only if it meets every success criterion: 1. Flesch Reading Ease >= 60 (scorer's own heuristic syllable count). 2. 180–350 words. 3. All 14 key rules/facts from the policy present (attendance threshold, reimbursement cap and period, extra chair allowance with medical note, claim deadline, itemised receipt, ownership threshold, return deadline, excluded categories internet/utilities/decor, approval time, payroll timing). 4. No new numbers: every number must appear in the source policy (integers 0–3 allowed). 5. No legalese words (notwithstanding, herein, pursuant, shall, aforementioned, enumerated). 6. Second person: 'you/your' used at least 5 times. 7. Contains a 'Quick example' section.
Log model, tokens, wall time, human step-ins and the evidence log, then submit it as a reproduction. A different model? Submit a head-to-head rerun.

How runs are scored: Methodology.

Was this useful?

⇄ Rerun on another model

Automated checks

Trust 80% comes from these checks (how Trust % works).

Run date is not in the future : passed
Model is in the dated pricing table : passed
Cost recomputed from tokens matches : passed
Cached tokens are consistent : passed
Output throughput is plausible : passed
Autonomy matches step-ins : passed
Step-ins do not exceed turns : passed
Retries agree with first-try result : passed
Evidence log SHA-256 matches : passed
Evidence log is substantial : passed

Evidence log

[
 {
  "duration_api_ms": 38527,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.156541,
  "usage": {
   "input_tokens": 10,
   "cache_creation_input_tokens": 9326,
   "cache_read_input_tokens": 88765,
   "output_tokens": 3207,
   "output_tokens_details": {
    "thinking_tokens": 1197
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 9326,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 615,
     "cache_read_input_tokens": 20569,
     "cache_creation_input_tokens": 903,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 903
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-opus-5-5": {
    "inputTokens": 10,
    "outputTokens": 3207,
    "cacheReadInputTokens": 88765,
    "cacheCreationInputTokens": 9326,
    "webSearchRequests": 0,
    "costUSD": 0.156541,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 1197,
    "canonicalModel": "claude-opus-5-5",
    "provider": "firstParty",
    "costBasis": "list"
   }
  },
  "permission_denials": [],
  "terminal_reason": "completed",
  "fast_mode_state": "off",
  "fast_mode_disabled_reason": "sdk_opt_in_required",
  "subagent_stats": {
   "spawned": 0,
   "requested": {
    "background": 0,
    "foreground": 0,
    "unset": 0
   },
   "started_in_background": 0,
   "max_depth": 0,
   "spawned_by_subagents": 0,
   "completed": 0,
   "failed": 0,
   "killed": {
    "parent": 0,
    "user": 0,
    "system": 0
   },
   "refused": {
    "depth_limit": 0,
    "concurrency_limit": 0,
    "budget": 0
   },
   "by_type": {}
  },
  "safety_stops": 0,
  "is_error": false,
  "num_turns": 5,
  "subtype": "success",
  "api_error_status": null,
  "result": "I wrote `policy_plain.md`. By my own count it scores about 85 on Flesch Reading Ease and runs about 316 words. Counting every space-separated token, including the `#` and `-` markers, gives 338, which is still under 350. I used a small script with an estimated syllable count because `textstat` isn't installed, so a different tool may give a slightly different score. It should still be well above 60.\n\n**How it's laid out:** a two-sentence intro, then short headed sections with bullets: who can use it, how much you can get, how to ask, who owns the items, what is not covered, and when you get paid. It ends with a \"Quick example\".\n\n**Rules and numbers:** all of them are kept: Rev. 4, fewer than 3 office days, USD 750 per rolling 24 months, an extra USD 300 for an ergonomic chair with medical papers, 60 days, the receipt contents, the USD 200 ownership line, the 14-day return for any reason, the four excluded types of spending, 10 business days, and the next payroll. \"Notwithstanding any prior arrangements\" became \"It replaces any past deals.\" The legalese check found no \"notwithstanding\", \"herein\", \"pursuant\" or \"shall\".\n\n**Quick example:** someone buys a USD 200 desk, sends the receipt within 60 days, gets paid in the next payroll, and must return the desk within 14 days if they leave. It only uses numbers from the policy, so I left out any subtraction, like the 750 cap minus the desk cost, because the result would be a new number.\n\n**Wording choices to review:**\n- \"Seating\" became \"chairs.\"\n- \"Medically recommended requirement\" became \"papers that show a medical need.\"\n- \"Deemed ineligible and shall not be processed\" became \"will not be paid.\"",
  "ttft_ms": 2713,
  "type": "result",
  "duration_ms": 38897,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 1697,
  "time_to_request_ms": 205,
  "first_content_frame_ms": 1697,
  "queued_turn_count": 0,
  "result_index": 0
 }
]