Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Cost Budget Check · ruvnet | DecimalAI

ruvnet/cost-budget-check

SkillScore86Maintained

Read accumulated cost-tracking spend + budget config, compute utilization, emit 50/75/90/100% alert ladder

agent-toolingby ruvnet

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

AI Engineer pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/ruvnet-cost-budget-check/SKILL.md

decimalai skills pull ruvnet-cost-budget-check --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/ruvnet-cost-budget-check/SKILL.md https://app.decimal.ai/s/ruvnet-cost-budget-check/SKILL.md

Raw file: /s/ruvnet-cost-budget-check/SKILL.md ↗ · machine-readable summary at /s/ruvnet-cost-budget-check.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 51d ago

Yes — +68% lift-25% tokens0% turns86% of whole cases pass with skill (19/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-02	✗→✓	▲ Improved	-38%	0%
case-11	✗→✓	▲ Improved	7%	0%
case-03	✗→✓	▲ Improved	-12%	0%
case-07	✗→✓	▲ Improved	-37%	0%
case-08	✗→✓	▲ Improved	-17%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/2/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

ruvnet / ruflo

plugins/ruflo-cost-tracker/skills/cost-budget-check/SKILL.md

View on GitHub ↗

License MITImported 8/2/2026Version v1

Skill contents

SKILL.md

Cost Budget Check

Reads cost-tracking:budget-config for the project's budget limit, sums total_cost_usd across session-* records produced by cost-track, computes utilization, and emits the measured 4-tier alert ladder (50% INFO / 75% WARNING / 90% CRITICAL / 100% HARD_STOP).

Until P2 (this skill) landed, the README documented the alert ladder but no code checked it. Now this skill is the gate.

When to use

After every cost-track run, to surface the alert level.
Before spawning a swarm — if utilization ≥ 90%, escalate to /cost-optimize first.
Cron-friendly via /loop 30m for continuous monitoring.

Steps

Run the check:

bash node plugins/ruflo-cost-tracker/scripts/budget.mjs check

Filter by period: BUDGET_PERIOD=today (default all). Use BUDGET_QUIET=1 for machine-readable JSON.

Inspect the markdown summary — budget, spent, remaining, utilization percentage, alert level (🟢 OK · 🟡 INFO · 🟠 WARNING · 🔴 CRITICAL · 🛑 HARD_STOP), and the recommended action.

Set / inspect the budget:

bash node plugins/ruflo-cost-tracker/scripts/budget.mjs set 50.00 node plugins/ruflo-cost-tracker/scripts/budget.mjs get

HARD_STOP path — budget.mjs check exits with code 1 when utilization ≥ 100%. Wrap critical agent spawns in a budget.mjs check && spawn … guard to fail closed.

Storage shape (`cost-tracking:budget-config`)

json
{
  "budget_usd": 50.00,
  "setAt": "2026-05-05T...",
  "thresholds": { "info": 0.50, "warning": 0.75, "critical": 0.90, "hard_stop": 1.00 }
}

Alert ladder (from REFERENCE.md, now enforced)

| Threshold | Level | Action | |---|---|---| | 50% | INFO 🟡 | log notification, no UX disruption | | 75% | WARNING 🟠 | display warning, suggest /cost-optimize | | 90% | CRITICAL 🔴 | urgent alert, recommend model downgrades | | 100% | HARD_STOP 🛑 | halt non-essential spawns; exit code 1 |

Cross-references

cost-track (producer) — populates cost-tracking:session-*
cost-report — same data source, narrative format
cost-optimize — recommended action when WARNING/CRITICAL
REFERENCE.md "Budget alert thresholds" — the documented ladder this enforces

Illustrative Want to see what ruvnet/cost-budget-check would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/7/2026, 20:50 UTC · 22 cases

Lift

+68%

86% vs 18%

Agent turns

0%

1.0 vs 1.0

Tokens

-26%

1,359.77 vs 1,825.95

fail → pass · 16pass → pass · 3fail → fail · 2pass → fail · 1

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 I'm about to launch a multi-agent swarm task and want to make sure we haven't hit any spending limits. Could you check our total accumulated session costs against our spending limit and give me a summary of our budget utilization and current alert tier?	fail→fail	7,566	2,497	-67%	1	1	0%	1,232	1,023	-17%	0	0	—
case-02 Can you run a check on today's resource expenditure against our budget allocation and output a clean JSON report showing total spent, remaining funds, and status?	fail→pass	7,639	1,797	-76%	1	1	0%	1,498	935	-38%	0	0	—
case-11 Our total usage has reached 80% of our spending limit. What alert level applies at 75%, and what optimization step should be suggested to the user?	fail→pass	5,805	2,127	-63%	1	1	0%	1,005	1,073	+7%	0	0	—
case-03 Please update our project spending limit to $50.00 and then verify our current accumulated costs against that new target to show me where we stand.	fail→pass	8,666	5,598	-35%	1	1	0%	1,063	939	-12%	0	0	—
case-04 We need to generate narrative summaries and executive commentary on our API spending trends over the past month. What CLI script or subcommand handles compiling these high-level textual cost breakdown reports?	pass→pass	13,642	3,438	-75%	1	1	0%	2,644	1,349	-49%	0	0	—
case-05 I want to construct a rule engine that automatically optimizes prompt tokens and model routing to lower API overhead. How do I configure cost optimization rules?	pass→fail	15,099	14,424	-4%	1	1	0%	2,783	3,758	+35%	0	0	—
case-06 How do I instrument my agent execution code to write individual session cost tracking records into storage?	fail→fail	18,220	10,839	-41%	1	1	0%	3,956	3,061	-23%	0	0	—
case-07 We want to check what our currently configured budget limit amount is without running a full session cost check or changing any settings. What CLI command displays the stored budget limit?	fail→pass	9,683	1,956	-80%	1	1	0%	1,676	1,051	-37%	0	0	—
case-08 When looking up budget parameters directly in our storage layer, under what storage key name is the budget configuration stored?	fail→pass	7,171	1,790	-75%	1	1	0%	1,092	902	-17%	0	0	—
case-09 What fields compose the JSON configuration stored under the budget storage key in our cost tracker?	fail→pass	4,593	3,951	-14%	1	1	0%	878	1,575	+79%	0	0	—
case-10 Our session costs have just reached 55% of the total budget. What alert level is triggered at the 50% mark, and what is the required operational response?	fail→pass	10,347	2,823	-73%	1	1	0%	1,781	1,255	-30%	0	0	—
case-12 Accumulated session costs have hit 92% of our budget limit. What alert tier is active at 90%, and what concrete action is recommended?	pass→pass	6,464	3,909	-40%	1	1	0%	1,091	1,370	+26%	0	0	—
case-13 Our expenditure has reached 102% of the configured spending threshold. What alert level is triggered at 100%, and what happens to non-essential agent spawns?	fail→pass	6,342	2,689	-58%	1	1	0%	982	1,155	+18%	0	0	—
case-14 When the spending check script executes and utilization is at or above 100%, what process exit code does it return to block downstream automated steps?	pass→pass	4,937	1,716	-65%	1	1	0%	843	914	+8%	0	0	—
case-15 How should critical agent spawn shell commands be guarded to prevent execution when the budget is exhausted?	fail→pass	15,694	7,100	-55%	1	1	0%	2,751	2,013	-27%	0	0	—
case-16 We want to run periodic continuous budget checks in our background task runner every 30 minutes. What command loop construct should be used?	fail→pass	9,724	3,102	-68%	1	1	0%	1,764	1,296	-27%	0	0	—
case-17 If no period environment variable is specified when invoking the budget check CLI script, what is the default evaluation period used when aggregating session records?	fail→pass	5,508	1,131	-79%	1	1	0%	870	874	+0%	0	0	—
case-18 Which record key prefix does the budget script aggregate from cost tracking to calculate total spent USD?	fail→pass	11,095	1,580	-86%	1	1	0%	1,783	950	-47%	0	0	—
case-19 Which numerical field name within each session record is summed up to compute total spent USD?	fail→pass	4,975	1,405	-72%	1	1	0%	813	915	+13%	0	0	—
case-20 We are preparing to launch an autonomous agent swarm. At what spending utilization percentage threshold must we stop and escalate to cost optimization prior to launching?	fail→pass	12,105	2,617	-78%	1	1	0%	2,005	1,220	-39%	0	0	—
case-21 In the storage configuration JSON for thresholds, what are the four precise object keys mapping to the decimal threshold levels?	fail→pass	25,610	2,537	-90%	1	1	0%	4,948	1,170	-76%	0	0	—
case-22 In the storage configuration JSON, what numerical decimal values correspond to the four budget threshold levels?	fail→pass	15,627	2,449	-84%	1	1	0%	2,713	1,117	-59%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +68 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 18.2%
Pass rate with the skill: 86.4%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 1
Skill version measured: v1
Run completed: 8/7/2026, 20:50 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-02

Prompt

Can you run a check on today's resource expenditure against our budget allocation and output a clean JSON report showing total spent, remaining funds, and status?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run86

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns-21% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 51 days ago

Other measured skills in the registry, with their headline benchmark lift.