aicraftbook/aicraftbook-skills / skills/cost-compass
cost-compass
VerifiedEstimates the monthly running cost of an AI feature, agent or automation from a description of its workload. Breaks the cost into model calls, input, output and cached tokens, and other paid services (search, embeddings, vector database, SMS, voice minutes, automation platform), shows every assumption and the arithmetic, gives a low / expected / high range, and names the two or three changes that would cut the cost most. Uses only prices the user gives or confirms from the provider's pricing page and never invents prices. Use when someone asks "how much will this cost to run", "estimate my API bill", "what will this agent cost per month", "is this automation affordable at 10,000 runs", or wants to compare the running cost of two designs or models.
| File | Kind | Size |
|---|---|---|
| CHANGELOG.md | markdown | 1.5 KB |
| LICENSE | other | 1.0 KB |
| README.md | markdown | 10.5 KB |
| SKILL.md | skill md | 6.6 KB |
| evals/independent-2026-10-21.json | config | 60.7 KB |
| evals/independent-2026-10-22.json | config | 54.5 KB |
| evals/results-2026-10-22.json | config | 5.4 KB |
| evals/tasks.json | config | 6.0 KB |
| examples/requests.md | markdown | 1.3 KB |
| references/cost-model.md | markdown | 10.0 KB |
| references/worked-examples.md | markdown | 7.4 KB |
| skill.json | config | 2.0 KB |
evals/independent-2026-10-22.json
View on GitHub{
"skill": "cost-compass",
"date": "2026-10-22",
"kind": "independent functional evaluation",
"frozenTasks": {
"path": "skills/cost-compass/evals/independent-runs/2026-10-22/frozen",
"sha256": "58577cabbd1d23d8868f4d91bec168c4b497821c9c9a9fe0b3c693168bd10d85",
"freezeRecord": "skills/cost-compass/evals/independent-runs/2026-10-22/FREEZE.txt"
},
"runner": {
"path": "web/scripts/eval-independent.ts",
"sha256": "2460924a28b6417066b1d41c9485d07e0d4fec0102341177695b178c9aa1a140"
},
"agentModel": "sonnet",
"graderModel": "opus",
"aggregates": {
"baseline": {
"n": 12,
"meanCombined": 0.766,
"meanRubric": 0.74,
"meanCheckFraction": 0.792,
"falseClaimRuns": 4,
"skillInvocationRate": 0,
"meanDurationMs": 17107.083,
"meanCostUsd": 0.08
},
"skill": {
"n": 12,
"meanCombined": 0.943,
"meanRubric": 0.885,
"meanCheckFraction": 1,
"falseClaimRuns": 2,
"skillInvocationRate": 0.833,
"meanDurationMs": 21123.417,
"meanCostUsd": 0.057
}
},
"perTask": {
"C1-rag-docs-assistant": {
"baseline": {
"n": 2,
"meanCombined": 0.907,
"meanRubric": 0.813,
"meanCheckFraction": 1,
"falseClaimRuns": 0,
"skillInvocationRate": 0,
"meanDurationMs": 20607.5,
"meanCostUsd": 0.037
},
"skill": {
"n": 2,
"meanCombined": 0.969,
"meanRubric": 0.938,
"meanCheckFraction": 1,
"falseClaimRuns": 0,
"skillInvocationRate": 1,
"meanDurationMs": 23229,
"meanCostUsd": 0.067
}
},
"C2-nightly-batch-summariser-cached": {
"baseline": {
"n": 2,
"meanCombined": 0.626,
"meanRubric": 0.875,
"meanCheckFraction": 0.375,
"falseClaimRuns": 0,
"skillInvocationRate": 0,
"meanDurationMs": 17157,
"meanCostUsd": 0.121
},
"skill": {
"n": 2,
"meanCombined": 0.938,
"meanRubric": 0.875,
"meanCheckFraction": 1,
"falseClaimRuns": 0,
"skillInvocationRate": 1,
"meanDurationMs": 21256,
"meanCostUsd": 0.058
}
},
"C3-agent-tool-calls-retries": {
"baseline": {
"n": 2,
"meanCombined": 0.907,
"meanRubric": 0.813,
"meanCheckFraction": 1,
"falseClaimRuns": 2,
"skillInvocationRate": 0,
"meanDurationMs": 14868.5,
"meanCostUsd": 0.04
},
"skill": {
"n": 2,
"meanCombined": 1,
"meanRubric": 1,
"meanCheckFraction": 1,
"falseClaimRuns": 1,
"skillInvocationRate": 1,
"meanDurationMs": 25852,
"meanCostUsd": 0.074
}
},
"E1-big-vs-small-router": {
"baseline": {
"n": 2,
"meanCombined": 0.907,
"meanRubric": 0.813,
"meanCheckFraction": 1,
"falseClaimRuns": 0,
"skillInvocationRate": 0,
"meanDurationMs": 15699,
"meanCostUsd": 0.037
},
"skill": {
"n": 2,
"meanCombined": 0.907,
"meanRubric": 0.813,
"meanCheckFraction": 1,
"falseClaimRuns": 1,
"skillInvocationRate": 1,
"meanDurationMs": 16838.5,
"meanCostUsd": 0.052
}
},
"M1-meeting-notes-no-prices-no-volume": {
"baseline": {
"n": 2,
"meanCombined": 0.438,
"meanRubric": 0.375,
"meanCheckFraction": 0.5,
"falseClaimRuns": 2,
"skillInvocationRate": 0,
"meanDurationMs": 18888.5,
"meanCostUsd": 0.205
},
"skill": {
"n": 2,
"meanCombined": 0.938,
"meanRubric": 0.875,
"meanCheckFraction": 1,
"falseClaimRuns": 0,
"skillInvocationRate": 1,
"meanDurationMs": 24346.5,
"meanCostUsd": 0.051
}
},
"T1-per-1k-mislabelled-as-per-1m": {
"baseline": {
"n": 2,
"meanCombined": 0.813,
"meanRubric": 0.75,
"meanCheckFraction": 0.875,
"falseClaimRuns": 0,
"skillInvocationRate": 0,
"meanDurationMs": 15422,
"meanCostUsd": 0.041
},
"skill": {
"n": 2,
"meanCombined": 0.907,
"meanRubric": 0.813,
"meanCheckFraction": 1,
"falseClaimRuns": 0,
"skillInvocationRate": 0,
"meanDurationMs": 15218.5,
"meanCostUsd": 0.039
}
}
},
"runs": [
{
"task": "C1-rag-docs-assistant",
"condition": "baseline",
"repeat": 1,
"blindId": "8388cc2ed1",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 29586,
"numTurns": 4,
"costUsd": 0.036554100000000006,
"tokens": {
"input": 6,
"cacheCreate": 6283,
"cacheRead": 22401,
"output": 917
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C1-total",
"pass": true,
"soft": false
},
{
"id": "C1-system-prompt-every-call",
"pass": true,
"soft": false
},
{
"id": "C1-output-line",
"pass": true,
"soft": false
},
{
"id": "C1-vector-db-line",
"pass": true,
"soft": false
},
{
"id": "C1-range",
"pass": false,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Uses 2,900 input tokens per call: the 800-token system prompt, 2,000 chunk tokens and 100 question tokens. Gives 145M tokens and $290."
},
{
"id": "R2",
"score": 1,
"reason": "Lists all six paid lines with shown arithmetic. Total is $459.15, matching the key."
},
{
"id": "R3",
"score": 1,
"reason": "States the chat model is about 94% ($430) and the rest about $29. Names fewer chunks and prompt caching as levers."
},
{
"id": "R4",
"score": 0.5,
"reason": "Uses only the test prices, but gives no low/expected/high range from varied assumptions."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.938
},
{
"task": "C1-rag-docs-assistant",
"condition": "baseline",
"repeat": 2,
"blindId": "50ae8e4c46",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 11629,
"numTurns": 4,
"costUsd": 0.037012500000000004,
"tokens": {
"input": 6,
"cacheCreate": 6273,
"cacheRead": 22385,
"output": 967
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C1-total",
"pass": true,
"soft": false
},
{
"id": "C1-system-prompt-every-call",
"pass": true,
"soft": false
},
{
"id": "C1-output-line",
"pass": true,
"soft": false
},
{
"id": "C1-vector-db-line",
"pass": true,
"soft": false
},
{
"id": "C1-range",
"pass": false,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Counts 800 + 2,000 + 100 = 2,900 tokens per call for all 50,000 calls: 145M tokens, $290."
},
{
"id": "R2",
"score": 1,
"reason": "All six paid lines are itemized with arithmetic shown, totalling $459.15."
},
{
"id": "R3",
"score": 0.5,
"reason": "Shows the chat model is about 94% of cost and the rest is minor, but names only prompt caching as a lever."
},
{
"id": "R4",
"score": 0.5,
"reason": "Uses only the test prices. Gives per-10k scaling but no low/expected/high range across stated assumptions."
}
],
"rubricScore": 0.75,
"falseClaims": [],
"combined": 0.875
},
{
"task": "C1-rag-docs-assistant",
"condition": "skill",
"repeat": 1,
"blindId": "ff33410c6b",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 0,
"durationMs": 28376,
"numTurns": 5,
"costUsd": 0.0588465,
"tokens": {
"input": 4,
"cacheCreate": 8567,
"cacheRead": 13305,
"output": 2324
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C1-total",
"pass": true,
"soft": false
},
{
"id": "C1-system-prompt-every-call",
"pass": true,
"soft": false
},
{
"id": "C1-output-line",
"pass": true,
"soft": false
},
{
"id": "C1-vector-db-line",
"pass": true,
"soft": false
},
{
"id": "C1-range",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Counts 800 + 5×400 + 100 = 2,900 tokens on every call, giving 145M input tokens and $290."
},
{
"id": "R2",
"score": 1,
"reason": "All six paid lines are itemized with arithmetic shown, totaling $459.15, which matches the key."
},
{
"id": "R3",
"score": 1,
"reason": "Says vector DB plus embeddings are about 6%, implying chat is about 94%. Names fewer chunks, answer cap and system-prompt caching as levers."
},
{
"id": "R4",
"score": 1,
"reason": "Low/expected/high range is driven by traffic, chunk size and output length, labeled as assumptions. Recomputed $320.72 and $699.30 correctly. Uses only test prices and flags the missing cache price."
}
],
"rubricScore": 1,
"falseClaims": [],
"combined": 1
},
{
"task": "C1-rag-docs-assistant",
"condition": "skill",
"repeat": 2,
"blindId": "43e8bd7f33",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 1,
"durationMs": 18082,
"numTurns": 7,
"costUsd": 0.0745547,
"tokens": {
"input": 6,
"cacheCreate": 12917,
"cacheRead": 25047,
"output": 2037
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C1-total",
"pass": true,
"soft": false
},
{
"id": "C1-system-prompt-every-call",
"pass": true,
"soft": false
},
{
"id": "C1-output-line",
"pass": true,
"soft": false
},
{
"id": "C1-vector-db-line",
"pass": true,
"soft": false
},
{
"id": "C1-range",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Counts 800 + 5×400 + 100 = 2,900 tokens on every call, giving 145M tokens and $290.00."
},
{
"id": "R2",
"score": 1,
"reason": "Includes all six paid lines with arithmetic shown, totalling $459.15, which matches the key."
},
{
"id": "R3",
"score": 1,
"reason": "Line table shows chat at $430 dominating and fixed costs as small. Names quantified levers: fewer chunks, answer cap, system-prompt caching."
},
{
"id": "R4",
"score": 0.5,
"reason": "Uses only the test prices, but explicitly declines to give a low/expected/high range."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.938
},
{
"task": "C2-nightly-batch-summariser-cached",
"condition": "baseline",
"repeat": 1,
"blindId": "5e590b66f7",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 20867,
"numTurns": 5,
"costUsd": 0.2052433,
"tokens": {
"input": 4,
"cacheCreate": 47287,
"cacheRead": 13073,
"output": 1478
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C2-total",
"pass": true,
"soft": false
},
{
"id": "C2-uncached-input-line",
"pass": true,
"soft": false
},
{
"id": "C2-output-line",
"pass": false,
"soft": false
},
{
"id": "C2-cache-read-line",
"pass": false,
"soft": false
},
{
"id": "C2-no-cache-comparison",
"pass": true,
"soft": true
}
],
"checkFraction": 0.5,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Separates cache write, cache reads and uncached transcripts, each at its own batch rate. Shown per night; monthly totals match the key."
},
{
"id": "R2",
"score": 1,
"reason": "Applies 50% to every line including write and reads; $15.986/night × 30 = $479.59, with workings shown."
},
{
"id": "R3",
"score": 1,
"reason": "Without caching $843.75/month; saving $364.16 (~43%), broken down per night and per month."
},
{
"id": "R4",
"score": 0.5,
"reason": "Notes transcripts (~$270) and output (~$169) dominate, but names no levers and gives no low/high range."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.688
},
{
"task": "C2-nightly-batch-summariser-cached",
"condition": "baseline",
"repeat": 2,
"blindId": "f878cc710c",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 13447,
"numTurns": 3,
"costUsd": 0.0371178,
"tokens": {
"input": 4,
"cacheCreate": 6183,
"cacheRead": 13078,
"output": 1107
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C2-total",
"pass": true,
"soft": false
},
{
"id": "C2-uncached-input-line",
"pass": false,
"soft": false
},
{
"id": "C2-output-line",
"pass": false,
"soft": false
},
{
"id": "C2-cache-read-line",
"pass": false,
"soft": false
},
{
"id": "C2-no-cache-comparison",
"pass": true,
"soft": true
}
],
"checkFraction": 0.25,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Separates uncached transcripts, cache reads, and one cache write per night, each at its own rate; nightly figures scale to 180M/270M/180K."
},
{
"id": "R2",
"score": 1,
"reason": "Applies 50% batch discount to all lines; nightly table shown; arrives at $479.59/month."
},
{
"id": "R3",
"score": 1,
"reason": "Without caching $843.75/month; saves $364.16, roughly 43%; arithmetic shown."
},
{
"id": "R4",
"score": 0.5,
"reason": "Says output and transcripts dominate remaining spend, but names no cost levers and gives no low/high range."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.563
},
{
"task": "C2-nightly-batch-summariser-cached",
"condition": "skill",
"repeat": 1,
"blindId": "4466189583",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 0,
"durationMs": 20254,
"numTurns": 5,
"costUsd": 0.0573178,
"tokens": {
"input": 4,
"cacheCreate": 8582,
"cacheRead": 13318,
"output": 2165
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C2-total",
"pass": true,
"soft": false
},
{
"id": "C2-uncached-input-line",
"pass": true,
"soft": false
},
{
"id": "C2-output-line",
"pass": true,
"soft": false
},
{
"id": "C2-cache-read-line",
"pass": true,
"soft": false
},
{
"id": "C2-no-cache-comparison",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Splits into 180M uncached, 270M cache reads and 180,000 cache writes, each priced at its own rate."
},
{
"id": "R2",
"score": 1,
"reason": "Applies 50% to every line including cache write and read; shows arithmetic; total $479.59 ($479.5875)."
},
{
"id": "R3",
"score": 1,
"reason": "Without caching $843.75; saves $364.16/month, 43% of bill and 90% of prefix cost."
},
{
"id": "R4",
"score": 0.5,
"reason": "Names transcript as largest line; levers are trimming transcripts and capping output. Omits fewer examples and declines a low/high range."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.938
},
{
"task": "C2-nightly-batch-summariser-cached",
"condition": "skill",
"repeat": 2,
"blindId": "15a9e1bf75",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 0,
"durationMs": 22258,
"numTurns": 5,
"costUsd": 0.05892080000000001,
"tokens": {
"input": 4,
"cacheCreate": 8563,
"cacheRead": 13308,
"output": 2333
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C2-total",
"pass": true,
"soft": false
},
{
"id": "C2-uncached-input-line",
"pass": true,
"soft": false
},
{
"id": "C2-output-line",
"pass": true,
"soft": false
},
{
"id": "C2-cache-read-line",
"pass": true,
"soft": false
},
{
"id": "C2-no-cache-comparison",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Separates cache writes (180,000), cache reads (270M) and uncached transcript input (180M), each priced at its own rate."
},
{
"id": "R2",
"score": 1,
"reason": "Applies 50% to every line including cache reads and writes; shows the arithmetic and reaches $479.59."
},
{
"id": "R3",
"score": 1,
"reason": "Gives $843.75 without caching, a saving of $364.16 a month, and 43%."
},
{
"id": "R4",
"score": 0.5,
"reason": "Names output and transcripts as main drivers with trimming levers. Omits fewer worked examples and explicitly gives no low/high range."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.938
},
{
"task": "C3-agent-tool-calls-retries",
"condition": "baseline",
"repeat": 1,
"blindId": "2a0d62bcad",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 13533,
"numTurns": 4,
"costUsd": 0.0402495,
"tokens": {
"input": 6,
"cacheCreate": 6339,
"cacheRead": 22415,
"output": 1264
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C3-total",
"pass": true,
"soft": false
},
{
"id": "C3-history-growth",
"pass": true,
"soft": false
},
{
"id": "C3-search-line",
"pass": true,
"soft": false
},
{
"id": "C3-retries-included",
"pass": true,
"soft": false
},
{
"id": "C3-budget-verdict",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Correctly models history growth: 3,200/5,000/6,800/8,600 = 23,600 input tokens per run."
},
{
"id": "R2",
"score": 1,
"reason": "Includes 8,800 runs and 17,600 searches ($176), with clear arithmetic arriving at $436.48."
},
{
"id": "R3",
"score": 0.5,
"reason": "Says it fits with ~13% headroom, but the sensitivity figures are wrong. Omits tool-result size and turn count as risks."
},
{
"id": "R4",
"score": 0.5,
"reason": "Identifies search (~40%) and input as the biggest costs and mentions caching. Doesn't propose trimming tool results, cutting retries or reducing searches."
}
],
"rubricScore": 0.75,
"falseClaims": [
"At 8,000 tasks the break-even retry rate is about 24% (actually ~26%: 500/(8,000×0.0496)=1.26)",
"At a 10% retry rate volume could grow to roughly 10,200 tasks (actually ~9,160: 500/(1.1×0.0496))",
"Headroom covers about 13% more volume (actually ~14.6%: 63.52/436.48)"
],
"combined": 0.875
},
{
"task": "C3-agent-tool-calls-retries",
"condition": "baseline",
"repeat": 2,
"blindId": "3301bb90ae",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 16204,
"numTurns": 4,
"costUsd": 0.0398951,
"tokens": {
"input": 6,
"cacheCreate": 6333,
"cacheRead": 22411,
"output": 1231
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C3-total",
"pass": true,
"soft": false
},
{
"id": "C3-history-growth",
"pass": true,
"soft": false
},
{
"id": "C3-search-line",
"pass": true,
"soft": false
},
{
"id": "C3-retries-included",
"pass": true,
"soft": false
},
{
"id": "C3-budget-verdict",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Correctly models history growth: 3,200/5,000/6,800/8,600 = 23,600 input tokens per run, with the reasoning explained."
},
{
"id": "R2",
"score": 1,
"reason": "Uses 8,800 runs and 2 searches per run ($176), and reaches $436.48 with per-run and monthly arithmetic shown."
},
{
"id": "R3",
"score": 1,
"reason": "Says it fits with about $64 (13%) headroom and that ~15% volume growth or ~25% re-run rate would push it over $500."
},
{
"id": "R4",
"score": 0.5,
"reason": "Gives input (48%) and search (40%) shares and suggests prompt caching. Omits trimming tool results and cutting retries or searches as levers."
}
],
"rubricScore": 0.875,
"falseClaims": [
"The search tool's added tokens are called 'small', but each search returns a 1,500-token result that is re-sent on later turns. Search results account for about 7,500 of the 23,600 input tokens per run, roughly $66/month.",
"The ~25% re-run threshold is slightly low: at 25% the total is about $496, still under $500. The break-even rate is about 26%."
],
"combined": 0.938
},
{
"task": "C3-agent-tool-calls-retries",
"condition": "skill",
"repeat": 1,
"blindId": "a7da45b5f0",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 0,
"durationMs": 24041,
"numTurns": 5,
"costUsd": 0.0649437,
"tokens": {
"input": 4,
"cacheCreate": 8626,
"cacheRead": 13317,
"output": 2910
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C3-total",
"pass": true,
"soft": false
},
{
"id": "C3-history-growth",
"pass": true,
"soft": false
},
{
"id": "C3-search-line",
"pass": true,
"soft": false
},
{
"id": "C3-retries-included",
"pass": true,
"soft": false
},
{
"id": "C3-budget-verdict",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Per-turn inputs of 3,200/5,000/6,800/8,600 are shown and summed to 23,600 per run, with re-sent history modeled correctly."
},
{
"id": "R2",
"score": 1,
"reason": "Uses 8,800 runs including retries and 17,600 searches costing $176. Arithmetic is shown and totals $436.48."
},
{
"id": "R3",
"score": 1,
"reason": "Says it fits the expected case, gives a $370–$571 range, and names re-run rate, tool size and output as drivers that could exceed budget."
},
{
"id": "R4",
"score": 1,
"reason": "Names input ($207.68) and search ($176) as top lines. Covers caching, trimming tool results, cutting retries and cutting searches, with correct savings."
}
],
"rubricScore": 1,
"falseClaims": [],
"combined": 1
},
{
"task": "C3-agent-tool-calls-retries",
"condition": "skill",
"repeat": 2,
"blindId": "96da801d09",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 1,
"durationMs": 27663,
"numTurns": 7,
"costUsd": 0.08357010000000001,
"tokens": {
"input": 6,
"cacheCreate": 12973,
"cacheRead": 25061,
"output": 2916
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "C3-total",
"pass": true,
"soft": false
},
{
"id": "C3-history-growth",
"pass": true,
"soft": false
},
{
"id": "C3-search-line",
"pass": true,
"soft": false
},
{
"id": "C3-retries-included",
"pass": true,
"soft": false
},
{
"id": "C3-budget-verdict",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Models history growth correctly: 12,800 static plus 10,800 history gives 23,600 input tokens per run, not 4 x 3,200."
},
{
"id": "R2",
"score": 1,
"reason": "Counts 8,800 runs from 10% re-runs and 17,600 searches costing $176. Total is $436.48 with line-by-line arithmetic shown."
},
{
"id": "R3",
"score": 1,
"reason": "Says it fits with $63.52 (~13%) headroom. Gives break-even thresholds for retry rate (~26%) and volume (~9,160 tasks)."
},
{
"id": "R4",
"score": 1,
"reason": "Names input ($207.68) and search ($176) as top costs. Proposes prefix caching, resuming retries and trimming tool results."
}
],
"rubricScore": 1,
"falseClaims": [
"Trimming savings omits that tool result 3 is re-sent on turn 4: cutting each result by 500 tokens saves 8,800 x 500 x 6 / 1M = $26.40, not $22.00.",
"A6 labels re-runs including searches as an assumption, but the design notes state it explicitly."
],
"combined": 1
},
{
"task": "E1-big-vs-small-router",
"condition": "baseline",
"repeat": 1,
"blindId": "23758b5912",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 16283,
"numTurns": 3,
"costUsd": 0.0359497,
"tokens": {
"input": 4,
"cacheCreate": 6076,
"cacheRead": 13077,
"output": 1033
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "E1-design-a",
"pass": true,
"soft": false
},
{
"id": "E1-design-b",
"pass": true,
"soft": false
},
{
"id": "E1-b-cheaper",
"pass": true,
"soft": false
},
{
"id": "E1-savings",
"pass": true,
"soft": false
},
{
"id": "E1-break-even",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Per-call input/output arithmetic shown; A = $1,500 and B = $75 + $300 = $375, all correct."
},
{
"id": "R2",
"score": 1,
"reason": "States B is cheaper by $1,125 (75%), with the small model explicitly charged on every review."
},
{
"id": "R3",
"score": 1,
"reason": "Break-even at 95% derived from the per-review formula; gives $15 per point and $675 at 40%."
},
{
"id": "R4",
"score": 0.5,
"reason": "Mentions small-model accuracy and flag-rate risk, but not confidence-flag calibration explicitly or escalation latency."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.938
},
{
"task": "E1-big-vs-small-router",
"condition": "baseline",
"repeat": 2,
"blindId": "10274724de",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 15115,
"numTurns": 4,
"costUsd": 0.0377651,
"tokens": {
"input": 6,
"cacheCreate": 6178,
"cacheRead": 22411,
"output": 1080
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "E1-design-a",
"pass": true,
"soft": false
},
{
"id": "E1-design-b",
"pass": true,
"soft": false
},
{
"id": "E1-b-cheaper",
"pass": true,
"soft": false
},
{
"id": "E1-savings",
"pass": true,
"soft": false
},
{
"id": "E1-break-even",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Correct per-call costs ($0.003+$0.002 input/output) give A=$1,500; B=$75 small plus $300 escalations=$375. Arithmetic shown."
},
{
"id": "R2",
"score": 1,
"reason": "States $1,125/month saving (75%) and that every review pays for the small model, including escalated ones."
},
{
"id": "R3",
"score": 0.5,
"reason": "Correct 95% break-even with formula; notes 100% overshoot of $75. No intermediate sensitivity example such as 40% -> $675."
},
{
"id": "R4",
"score": 0.5,
"reason": "Flags only small-model tagging quality; omits confidence-flag calibration and extra latency on escalated reviews."
}
],
"rubricScore": 0.75,
"falseClaims": [],
"combined": 0.875
},
{
"task": "E1-big-vs-small-router",
"condition": "skill",
"repeat": 1,
"blindId": "65d73dd10e",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 0,
"durationMs": 17151,
"numTurns": 5,
"costUsd": 0.052973599999999996,
"tokens": {
"input": 4,
"cacheCreate": 8471,
"cacheRead": 13316,
"output": 1775
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "E1-design-a",
"pass": true,
"soft": false
},
{
"id": "E1-design-b",
"pass": true,
"soft": false
},
{
"id": "E1-b-cheaper",
"pass": true,
"soft": false
},
{
"id": "E1-savings",
"pass": true,
"soft": false
},
{
"id": "E1-break-even",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Correct line-by-line arithmetic: A $900+$600=$1,500; B $45+$30 small on all, $180+$120 escalations, totalling $375."
},
{
"id": "R2",
"score": 0.5,
"reason": "States $1,125/month (75%) saving. Shows small model applied to all 180M tokens but never explicitly says escalated reviews pay both passes."
},
{
"id": "R3",
"score": 1,
"reason": "Gives B = $75 + f×$1,500, break-even at 95%, and a sensitivity table at 20/50/80/95%."
},
{
"id": "R4",
"score": 0.5,
"reason": "Flags small-model quality and confidently-wrong cases (calibration), but does not mention latency on escalated reviews."
}
],
"rubricScore": 0.75,
"falseClaims": [],
"combined": 0.875
},
{
"task": "E1-big-vs-small-router",
"condition": "skill",
"repeat": 2,
"blindId": "0619327028",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 0,
"durationMs": 16526,
"numTurns": 5,
"costUsd": 0.051021300000000006,
"tokens": {
"input": 4,
"cacheCreate": 8468,
"cacheRead": 13313,
"output": 1581
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "E1-design-a",
"pass": true,
"soft": false
},
{
"id": "E1-design-b",
"pass": true,
"soft": false
},
{
"id": "E1-b-cheaper",
"pass": true,
"soft": false
},
{
"id": "E1-savings",
"pass": true,
"soft": false
},
{
"id": "E1-break-even",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Both totals are correct ($1,500 and $375). Per-call input and output arithmetic is shown, so the $900/$600 split is implicit rather than stated."
},
{
"id": "R2",
"score": 1,
"reason": "States $1,125/month saving and 75% less. Explicitly charges the small model on all 300,000 reviews, escalated ones included."
},
{
"id": "R3",
"score": 1,
"reason": "Gives per-review formula $0.00025 + p×$0.005 with break-even at 95%. Sensitivity example: 45% gives $750, which is correct."
},
{
"id": "R4",
"score": 0.5,
"reason": "Covers small-model quality and confidence-flag reliability. Does not mention latency on escalated reviews."
}
],
"rubricScore": 0.875,
"falseClaims": [
"A5 labels 'no retry cost' as given, but neither file mentions retries; only caching and batch discounts are excluded."
],
"combined": 0.938
},
{
"task": "M1-meeting-notes-no-prices-no-volume",
"condition": "baseline",
"repeat": 1,
"blindId": "425ad217e6",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 20682,
"numTurns": 4,
"costUsd": 0.2047836,
"tokens": {
"input": 6,
"cacheCreate": 46763,
"cacheRead": 22496,
"output": 1547
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "M1-no-invented-token-price",
"pass": true,
"soft": false
},
{
"id": "M1-no-invented-monthly-total",
"pass": false,
"soft": false
},
{
"id": "M1-asks-volume",
"pass": true,
"soft": false
},
{
"id": "M1-asks-prices",
"pass": false,
"soft": false
},
{
"id": "M1-transcript-length",
"pass": true,
"soft": true
}
],
"checkFraction": 0.5,
"rubricItems": [
{
"id": "R1",
"score": 0,
"reason": "Gives a headline '$100/month, under $250' and specific per-MTok prices it cannot verify. It never says plainly that no number is possible."
},
{
"id": "R2",
"score": 0.5,
"reason": "Asks for meetings per month and offers to measure real transcripts. It doesn't ask which model or to confirm prices from the pricing page, or summary length."
},
{
"id": "R3",
"score": 0.5,
"reason": "Gives per-meeting token assumptions and a scenario table, which is somewhat usable. There is no explicit fill-in formula or worksheet."
},
{
"id": "R4",
"score": 0.5,
"reason": "Flags transcription, email, and longer meetings raising cost. It doesn't mention chunking or multiple calls for long transcripts."
}
],
"rubricScore": 0.375,
"falseClaims": [
"Presents specific Anthropic list prices ($0.10/$0.50, $2/$10, $4/$20 per MTok) as current fact; these are unverified and likely invented",
"Headline '$100/month, almost certainly under $250' rests on an invented 2,000 meetings/month and can't be supported"
],
"combined": 0.438
},
{
"task": "M1-meeting-notes-no-prices-no-volume",
"condition": "baseline",
"repeat": 2,
"blindId": "fc53ecc990",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 17095,
"numTurns": 4,
"costUsd": 0.2047084,
"tokens": {
"input": 6,
"cacheCreate": 46782,
"cacheRead": 22484,
"output": 1532
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "M1-no-invented-token-price",
"pass": true,
"soft": false
},
{
"id": "M1-no-invented-monthly-total",
"pass": false,
"soft": false
},
{
"id": "M1-asks-volume",
"pass": true,
"soft": false
},
{
"id": "M1-asks-prices",
"pass": false,
"soft": false
},
{
"id": "M1-transcript-length",
"pass": true,
"soft": true
}
],
"checkFraction": 0.5,
"rubricItems": [
{
"id": "R1",
"score": 0,
"reason": "Presents specific $/MTok rates as current list prices and gives a headline $30–$60/month total, even though volume and prices are unknown."
},
{
"id": "R2",
"score": 0.5,
"reason": "Asks for meeting count and whether transcripts exist. Does not ask for the chosen model, confirmed prices, transcript length or summary length."
},
{
"id": "R3",
"score": 0.5,
"reason": "Lists assumptions and a volume-sensitivity table but gives no explicit fill-in formula or worksheet the user can complete."
},
{
"id": "R4",
"score": 0.5,
"reason": "Flags transcription and email costs clearly. Notes long meetings and retries but not chunking or multiple calls per transcript."
}
],
"rubricScore": 0.375,
"falseClaims": [
"Per-1M-token rates for Opus 5.5 ($4/$20), Sonnet 5.5 ($2/$10) and Haiku 5.5 ($0.10/$0.50) are presented as current list prices without verification.",
"Haiku per-meeting cost of ~$0.002 understates its own inputs, which give ~$0.003 (~$1.40 at 500 meetings).",
"SendGrid is said to cost about $0.10 per 1,000 emails; that is SES pricing, not SendGrid's plan pricing."
],
"combined": 0.438
},
{
"task": "M1-meeting-notes-no-prices-no-volume",
"condition": "skill",
"repeat": 1,
"blindId": "613e9f1cf0",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 0,
"durationMs": 29820,
"numTurns": 4,
"costUsd": 0.05273950000000001,
"tokens": {
"input": 4,
"cacheCreate": 8001,
"cacheRead": 13275,
"output": 1940
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "M1-no-invented-token-price",
"pass": true,
"soft": false
},
{
"id": "M1-no-invented-monthly-total",
"pass": true,
"soft": false
},
{
"id": "M1-asks-volume",
"pass": true,
"soft": false
},
{
"id": "M1-asks-prices",
"pass": true,
"soft": false
},
{
"id": "M1-transcript-length",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Gives no prices or dollar total. Says plainly it can't give a figure without volume and model prices; price cells are left as placeholders."
},
{
"id": "R2",
"score": 1,
"reason": "Asks meetings/month, meeting length, chosen model and prices from the pricing page. Summary length is assumed (800) with a range rather than asked."
},
{
"id": "R3",
"score": 1,
"reason": "Gives a clear fill-in formula, a price sheet worksheet, token tables by volume and a low/expected/high range table."
},
{
"id": "R4",
"score": 0.5,
"reason": "Flags email cost. Never mentions transcription cost. Long transcripts are only an assumption (A6), with no chunking or multi-call warning."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.938
},
{
"task": "M1-meeting-notes-no-prices-no-volume",
"condition": "skill",
"repeat": 2,
"blindId": "d036f73d7e",
"skillAvailable": true,
"skillInvoked": true,
"skillFilesRead": 0,
"durationMs": 18873,
"numTurns": 4,
"costUsd": 0.048414200000000004,
"tokens": {
"input": 4,
"cacheCreate": 8012,
"cacheRead": 13282,
"output": 1503
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "M1-no-invented-token-price",
"pass": true,
"soft": false
},
{
"id": "M1-no-invented-monthly-total",
"pass": true,
"soft": false
},
{
"id": "M1-asks-volume",
"pass": true,
"soft": false
},
{
"id": "M1-asks-prices",
"pass": true,
"soft": false
},
{
"id": "M1-transcript-length",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Gives no dollar total or provider prices and says plainly it can't give a figure without volume and prices."
},
{
"id": "R2",
"score": 1,
"reason": "Asks for meetings per month, meeting length, and the model with its input and output prices. Summary length is assumed and stated, not asked."
},
{
"id": "R3",
"score": 1,
"reason": "Gives a fill-in formula and a cost-by-line worksheet with input, output, email and hosting lines, ready to fill in."
},
{
"id": "R4",
"score": 0.5,
"reason": "Flags email and reasoning tokens, but not transcription cost or chunking/multiple calls for long transcripts."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.938
},
{
"task": "T1-per-1k-mislabelled-as-per-1m",
"condition": "baseline",
"repeat": 1,
"blindId": "da33c8ccce",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 17126,
"numTurns": 4,
"costUsd": 0.0415718,
"tokens": {
"input": 6,
"cacheCreate": 6672,
"cacheRead": 22918,
"output": 1258
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "T1-unit-error-caught",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-current",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-tripled",
"pass": true,
"soft": false
},
{
"id": "T1-no-uncorrected-tripled",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-unit-price",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Identifies the per-1K quote versus per-1M column mismatch and the 1,000x understatement. Does not triple $1.872."
},
{
"id": "R2",
"score": 1,
"reason": "Gives $3/1M, $15/1M, $252 + $1,620 = $1,872, and $756 + $4,860 = $5,616 next quarter."
},
{
"id": "R3",
"score": 1,
"reason": "Raises the error alongside the headline number. Could not edit, offers specific corrected rows, and supplies a row to paste."
},
{
"id": "R4",
"score": 0,
"reason": "Does not mention output's ~87% cost share, the same-tokens-per-listing assumption, or the no-volume-discount assumption."
}
],
"rubricScore": 0.75,
"falseClaims": [],
"combined": 0.875
},
{
"task": "T1-per-1k-mislabelled-as-per-1m",
"condition": "baseline",
"repeat": 2,
"blindId": "e2080eae48",
"skillAvailable": false,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 13718,
"numTurns": 4,
"costUsd": 0.0400384,
"tokens": {
"input": 6,
"cacheCreate": 6581,
"cacheRead": 22924,
"output": 1141
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "T1-unit-error-caught",
"pass": false,
"soft": false
},
{
"id": "T1-corrected-current",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-tripled",
"pass": true,
"soft": false
},
{
"id": "T1-no-uncorrected-tripled",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-unit-price",
"pass": false,
"soft": true
}
],
"checkFraction": 0.75,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Identifies the per-1K quote vs per-1M column mismatch, says costs are 1000x too low, and does not triple $1.872."
},
{
"id": "R2",
"score": 1,
"reason": "Gives $3/$15 per 1M, $252+$1,620=$1,872 current, and $756+$4,860=$5,616 next quarter."
},
{
"id": "R3",
"score": 1,
"reason": "States the unit error right next to the requested number. It did not edit, citing a disabled tool, and offers to fix the sheet."
},
{
"id": "R4",
"score": 0,
"reason": "Does not mention the ~87% output share, the same-tokens-per-listing assumption, or the no-volume-discount assumption."
}
],
"rubricScore": 0.75,
"falseClaims": [],
"combined": 0.75
},
{
"task": "T1-per-1k-mislabelled-as-per-1m",
"condition": "skill",
"repeat": 1,
"blindId": "8d4e61519c",
"skillAvailable": true,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 15626,
"numTurns": 4,
"costUsd": 0.0427406,
"tokens": {
"input": 6,
"cacheCreate": 6967,
"cacheRead": 23406,
"output": 1252
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "T1-unit-error-caught",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-current",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-tripled",
"pass": true,
"soft": false
},
{
"id": "T1-no-uncorrected-tripled",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-unit-price",
"pass": true,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Explicitly identifies per-1K vs per-1M mismatch, 1000x understatement; gives $5,616 not $5.616."
},
{
"id": "R2",
"score": 1,
"reason": "States $3/1M and $15/1M, $1,872 current with 252/1620, and $5,616 next quarter with breakdown."
},
{
"id": "R3",
"score": 1,
"reason": "Error flagged in the opening paragraph alongside the number; couldn't edit, but gave corrected row values and offered fixes."
},
{
"id": "R4",
"score": 0,
"reason": "No mention of output share (~87%) or assumptions about constant tokens per listing and no volume discount."
}
],
"rubricScore": 0.75,
"falseClaims": [],
"combined": 0.875
},
{
"task": "T1-per-1k-mislabelled-as-per-1m",
"condition": "skill",
"repeat": 2,
"blindId": "dcd1d23e68",
"skillAvailable": true,
"skillInvoked": false,
"skillFilesRead": 0,
"durationMs": 14811,
"numTurns": 3,
"costUsd": 0.0361309,
"tokens": {
"input": 4,
"cacheCreate": 6215,
"cacheRead": 13329,
"output": 993
},
"exitCode": 0,
"timedOut": false,
"filesChanged": [],
"outsideWrites": [],
"permissionDenials": [],
"checks": [
{
"id": "T1-unit-error-caught",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-current",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-tripled",
"pass": true,
"soft": false
},
{
"id": "T1-no-uncorrected-tripled",
"pass": true,
"soft": false
},
{
"id": "T1-corrected-unit-price",
"pass": false,
"soft": true
}
],
"checkFraction": 1,
"rubricItems": [
{
"id": "R1",
"score": 1,
"reason": "Explicitly identifies the per-1K quote versus the per-1M column label, says costs are 1,000x too low, and avoids $5.616."
},
{
"id": "R2",
"score": 1,
"reason": "Gives $252, $1,620, $1,872 and $5,616 correctly. Prices stay per 1K instead of being restated as $3/$15 per 1M, which is equivalent."
},
{
"id": "R3",
"score": 1,
"reason": "Flags the error up front next to the number. It can't edit, so it gives exact cell fixes and the row to append."
},
{
"id": "R4",
"score": 0.5,
"reason": "States the same-tokens-per-listing and unchanged-price assumptions, but omits that output is about 87% of cost and doesn't mention volume discounts explicitly."
}
],
"rubricScore": 0.875,
"falseClaims": [],
"combined": 0.938
}
]
}Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |