Skip to content

Claude Sonnet 5.5 (effort default (no --effort flag) · thinking on) on Run a quarter-end leverage covenant test from the credit agreement definitions, restated quarters and mixed cash

AClaude Sonnet 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-21 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first try
Estimated cost
$0.05
Wall time
10s
Human step-ins
0
Outcome
Pass

Compare with other models on this task →

Summary

Exact prompt

# Run the quarter-end leverage covenant test Inputs: `agreement.json` (the credit agreement's definitions and covenant schedule), `quarterly.csv` (quarterly results; some quarters have more than one version) and `balance_sheet.json` (debt, leases and cash at the test date). Amounts are in USD thousands. Test the leverage covenant at the test date in `agreement.json`, applying the agreement's definitions exactly. - Use the most recently filed version of each quarter's numbers; forecasts are not results. - Consolidated EBITDA is over the four quarters ending on the test date. - Leverage = net debt / Consolidated EBITDA. The covenant is breached if leverage is above the maximum that applies on the test date. - EBITDA headroom = how far Consolidated EBITDA could fall before the covenant is breached, as % of Consolidated EBITDA: (EBITDA - net debt / maximum) / EBITDA x 100. Write `out/covenant.json`: `{"consolidated_ebitda": <2 decimals>, "net_debt": <2 decimals>, "leverage": <2 decimals>, "max_leverage": <x>, "breached": <true|false>, "ebitda_headroom_pct": <2 decimals>}`. Round half-up to 2 decimals for reporting only; decide the breach on unrounded figures. ## How this run works - Work on your own: nobody will answer questions in this task, and writing to `ask/` counts against the run. - Write only under `out/`. Create `out/` if it is missing. Inputs are synthetic; don't fetch anything from the network. - Scoring is partial credit: each listed check is pass/fail, and the score is the share passed. Numbers are compared to within 0.005 unless the task says otherwise.

Compute Consolidated EBITDA, net debt, leverage, breach and headroom exactly as a credit agreement defines them. Two of the four quarters were restated, a forecast quarter sits in the data, one-off add-backs are capped, leases count as debt, escrowed cash cannot be netted and cash netting is capped, and the maximum steps down mid-year. Odd seeds pass the covenant and even seeds breach, always at least 0.05x from the limit. 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). Folder-restore note: this run's output files under work/out/ were lost in a 1:22 AM ET restore on Oct 21, 2026. The trace, step-in exchanges and output-hash list were checked against the recorded SHA-256 values and match; the hashes of the individual output files inside that list can no longer be recomputed. Measured data unchanged.

Success criteria: ebitda net_debt leverage_and_limit breach headroom

Setup notes: Batch 19 case u07-leverage-covenant-test (offline, synthetic inputs, seed 62712241). Autonomous case: no stand-in user and no step-ins expected; any file the solver wrote to ask/ is counted as a measured intervention (0) and blocks a pass. Prompt = TASK.md verbatim. Hidden key kept outside the solver's work dir. Scored with the case's scorer (python3 score.py WORKDIR KEYDIR, absolute paths); pass = every check true and no ask/ files. Tools: Read/Write/Edit/Glob/Grep/Bash, web disabled. Single run, no retries. 1-min load average at scoring 0.26 (8 cores). Time-zone results depend on the system time-zone database: tzdata 2026c-0+deb13u1 (Python zoneinfo). Claude Code's own reported cost for claude-sonnet-5.5: $0.046056. Score: 5/5. 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
$0.05
Wall time
10s
Human step-ins
0
Turns
3
Model usage
ModelInput tokOutput tokCost
AClaude Sonnet 5.5anthropic/claude-sonnet-5.5effort default (no --effort flag) · thinking on58,0341,064$0.05

Reproduce this run

  1. Use the use case prompt on the use case page.
  2. Same setup: Claude Sonnet 5.5 (anthropic/claude-sonnet-5.5, effort default (no --effort flag) · thinking on) via Claude Code 2.1.299 (headless -p), cloud.
  3. It counts as a success only if it meets every success criterion: ebitda net_debt leverage_and_limit breach headroom
  4. 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).

Evidence log
[
 {
  "duration_api_ms": 8606,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.0460558,
  "usage": {
   "input_tokens": 6,
   "cache_creation_input_tokens": 7590,
   "cache_read_input_tokens": 50438,
   "output_tokens": 1064,
   "output_tokens_details": {
    "thinking_tokens": 390
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 7590,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 386,
     "cache_read_input_tokens": 19449,
     "cache_creation_input_tokens": 680,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 680
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-sonnet-5-5": {
    "inputTokens": 6,
    "outputTokens": 1064,
    "cacheReadInputTokens": 50438,
    "cacheCreationInputTokens": 7590,
    "webSearchRequests": 0,
    "costUSD": 0.0460558,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 390,
    "canonicalModel": "claude-sonnet-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": 3,
  "subtype": "success",
  "api_error_status": null,
  "result": "The covenant is not breached at the 2026-09-30 test date. Leverage is 2.04x against a maximum of 3.25x, and I wrote the result to `out/covenant.json`.\n\n- **Consolidated EBITDA: 64,096.80.** I used the latest filed version of each quarter from 2025Q4 to 2026Q3. That meant the restated 2025Q4 and the restated 2026Q3. Reported EBITDA is 53,414. One-off add-backs total 12,132, but the cap is 20% of reported EBITDA, or 10,682.80, so only that amount counts.\n- **Net debt: 130,916.** Total debt is 137,461: borrowings of 125,119 plus leases of 12,342. I netted only the 6,545 of operating cash, which is under the 10,000 cap. The 2,464 earn-out escrow is restricted, so it doesn't count.\n- **Maximum leverage: 3.25x.** The test date is after 2026-06-30, so the lower maximum applies.\n- **EBITDA headroom: 37.15%.**\n\nThe forecast row (2026Q4) is outside the four-quarter window and wasn't used.",
  "ttft_ms": 1778,
  "type": "result",
  "duration_ms": 8947,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 978,
  "time_to_request_ms": 262,
  "first_content_frame_ms": 978,
  "queued_turn_count": 0,
  "result_index": 0
 }
]

Get new runs and ranking changes weekly