Ready to supercharge your AI agents?

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

Get Started Free →

Attack Resilience Scoring · yogsoth-ai | DecimalAI

yogsoth-ai/attack-resilience-scoring

SkillScore57Maintained

Compute overall resilience score (0.0-1.0) based on attack results, coverage, and vulnerability severity distribution.

Securityby yogsoth-ai

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

Automatic

Security & AppSec pack

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

pip install decimalai

2 · Get the skill — writes .claude/skills/yogsoth-ai-attack-resilience-scoring/SKILL.md

decimalai skills pull yogsoth-ai-attack-resilience-scoring --out .claude/skills/

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

curl -fsSL --create-dirs -o .claude/skills/yogsoth-ai-attack-resilience-scoring/SKILL.md https://app.decimal.ai/s/yogsoth-ai-attack-resilience-scoring/SKILL.md

Raw file: /s/yogsoth-ai-attack-resilience-scoring/SKILL.md ↗ · machine-readable summary at /s/yogsoth-ai-attack-resilience-scoring.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

23 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 7d ago

Yes — +30% lift+18% tokens0% turns57% of whole cases pass with skill (13/23)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-07	✗→✓	▲ Improved	11%	0%
case-11	✗→✓	▲ Improved	41%	0%
case-14	✗→✓	▲ Improved	-1%	0%
case-15	✗→✓	▲ Improved	-34%	0%
case-17	✗→✓	▲ Improved	18%	0%

Showing 5 of 23 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

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

yogsoth-ai / de-anthropocentric-research-engine

skills/attack-resilience-scoring/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 8/3/2026Version v1

Skill contents

SKILL.md

Attack Resilience Scoring

Computes a quantitative resilience score for the artifact based on red team results.

Execution

Subagent — spawned via subagent-spawning/spawn-agent.

Why Subagent

Scoring requires calibrated judgment independent of attack or defense bias. The scorer must weigh findings objectively against coverage.

Input

aggregated_findings: Deduplicated vulnerability report from finding-aggregation
coverage_data: What percentage of threat surfaces were tested, at what depth

Output

resilience_score: 0.0-1.0 overall score
dimension_scores: Per-dimension breakdown (logical, empirical, methodological, practical)
confidence_in_score: How much to trust the score given coverage gaps
verdict: Pass/conditional-pass/fail with justification

Available SOPs

Optional, no fixed order; the final leaf is always a sop.

| SOP | When to use | | --- | --- | | spawn-agent | Spawn a customized CC subagent with full MCP tool access. Used by SOPs that declare execution: subagent. |

Want to see if yogsoth-ai/attack-resilience-scoring would activate on your phrasing? Try it before you install — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/22/2026, 06:44 UTC · 23 cases

Lift

+30%

57% vs 26%

Agent turns

0%

1.0 vs 1.0

Tokens

+18%

2,283.26 vs 1,929.09

fail → pass · 7pass → pass · 6fail → fail · 10

Per-case results (23)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 We just completed red-teaming our payment gateway. I'm providing the deduplicated vulnerabilities list from our aggregation step along with the surface testing coverage report. Please evaluate our system and output the overall 0-1 resilience metric, the breakdown across logical, empirical, methodological, and practical dimensions, our confidence level based on test completeness, and a pass/conditional-pass/fail determination with reasoning.	fail→fail	23,297	13,436	-42%	1	1	0%	2,484	2,580	+4%	0	0	—
case-02 I have attached our deduplicated findings from the latest pentest as well as the depth and coverage figures across our attack surfaces. Please generate a formal security assessment returning an overall score scaled 0 to 1, individual scores across the four dimensions (methodological, practical, empirical, logical), a confidence score based on test depth, and a pass or fail verdict accompanied by a clear justification.	fail→fail	6,162	16,511	+168%	1	1	0%	1,097	3,317	+202%	0	0	—
case-03 We ran multiple automated vulnerability scanners against our OAuth authentication service and collected 45 raw finding reports. Please deduplicate these findings by root cause, combine overlapping CVEs, and produce a unified aggregated vulnerability report.	pass→pass	6,550	11,276	+72%	1	1	0%	1,070	1,558	+46%	0	0	—
case-04 We need to test the security of our REST API gateway. Please execute active penetration testing payloads against the endpoint to discover SQL injection and SSRF vulnerabilities.	pass→pass	6,346	5,368	-15%	1	1	0%	605	769	+27%	0	0	—
case-05 Our vulnerability report shows a remote code execution issue in our Node.js express file upload middleware. Please write the patched code to sanitize incoming file paths and prevent directory traversal.	pass→pass	7,363	9,372	+27%	1	1	0%	1,412	2,034	+44%	0	0	—
case-06 We completed testing on AWS IAM architecture with 80% coverage. I have the deduplicated list of findings. Calculate our resilience metrics and pass/fail determination, but do it directly in this session to save tokens instead of spinning up new agents.	fail→fail	9,496	10,596	+12%	1	1	0%	1,457	2,056	+41%	0	0	—
case-07 Here are the aggregated vulnerabilities and surface depth data for our Kubernetes cluster red-team engagement. Generate the quantitative resilience evaluation. Break down the dimension metrics into technical risk, operational impact, and compliance.	fail→pass	22,137	23,136	+5%	1	1	0%	3,850	4,280	+11%	0	0	—
case-08 Evaluate the resilience of our fintech ledger API using our deduplicated vulnerability report and 95% surface coverage dataset. Express the overall resilience score as a percentage scale from 0% to 100%.	fail→fail	11,620	16,221	+40%	1	1	0%	2,107	3,103	+47%	0	0	—
case-09 Based on our red team summary and testing coverage data for the OAuth identity provider, provide a security status recommendation. Use the scale 'Approved', 'Rejected', or 'Needs Improvement'.	fail→fail	3,619	9,933	+174%	1	1	0%	578	1,858	+221%	0	0	—
case-10 We only tested 15% of our smart lock firmware attack surface at shallow depth before running out of time, but found no critical flaws in that 15%. Calculate our security score and assess confidence.	pass→pass	14,867	16,857	+13%	1	1	0%	2,378	3,133	+32%	0	0	—
case-11 Explain why our red team lead should not score our healthcare portal's resilience directly in the main conversation window, and initiate the appropriate workflow.	fail→pass	10,876	21,654	+99%	1	1	0%	1,683	2,372	+41%	0	0	—
case-12 Here is the aggregated vulnerability report for our e-commerce platform: [CVE-2023-1234 critical, CVE-2023-5678 high]. Compute our overall resilience score and dimension breakdown immediately.	fail→fail	10,178	9,463	-7%	1	1	0%	1,769	1,958	+11%	0	0	—
case-13 We achieved 90% attack surface coverage on our internal HR system. The aggregated findings reveal 3 medium-severity authorization flaws but zero critical vulnerabilities. Evaluate resilience and state the verdict.	pass→pass	10,606	8,988	-15%	1	1	0%	1,730	1,744	+1%	0	0	—
case-14 Red team findings for our core banking gateway show 2 critical unauthenticated remote code execution vulnerabilities across 95% surface coverage. Compute the scoring metrics and determination.	fail→pass	13,964	13,230	-5%	1	1	0%	2,435	2,404	-1%	0	0	—
case-15 Our red-team engagement covered 98% of attack surfaces at deep penetration levels with 0 high/critical findings and 1 low informational finding on the customer portal. Score our resilience.	fail→pass	13,951	7,495	-46%	1	1	0%	2,159	1,427	-34%	0	0	—
case-16 We have aggregated findings and 85% coverage data for our 5G core network equipment. Calculate the final resilience score and output object using your default local context.	fail→fail	7,030	15,887	+126%	1	1	0%	1,233	2,428	+97%	0	0	—
case-17 Evaluate the resilience of our CI/CD pipeline artifact repository using the provided deduplicated findings and 70% coverage metrics. Ensure all four standard dimension scores are populated in the JSON output.	fail→pass	13,582	16,523	+22%	1	1	0%	2,329	2,757	+18%	0	0	—
case-18 Compute the resilience evaluation for the infusion pump firmware given 88% attack coverage and 2 medium finding aggregations. Output the main score scaled to 10 points.	fail→pass	9,645	14,476	+50%	1	1	0%	1,775	2,705	+52%	0	0	—
case-19 Red team assessment for our zero-trust network agent shows moderate coverage with 1 unpatched high severity flaw. Return only a single word 'FAIL' for the final verdict without any text.	fail→fail	3,657	4,047	+11%	1	1	0%	212	668	+215%	0	0	—
case-20 The automated pentest on our autonomous drone software covered 100% of API endpoints, but only at surface-level fuzzing without deep stateful testing. Produce the evaluation report.	pass→pass	16,947	14,511	-14%	1	1	0%	2,745	2,663	-3%	0	0	—
case-21 Assess the attack resilience of our multi-sig wallet backend based on the aggregated vulnerability list and 92% coverage matrix provided. Execute this evaluation directly without calling external subagent SOPs.	fail→fail	22,334	17,104	-23%	1	1	0%	3,407	3,110	-9%	0	0	—
case-22 Analyze our PostgreSQL cluster red team results (90% depth/coverage, 2 low findings). Provide the resilience assessment following the required output contract structure.	fail→pass	33,260	9,304	-72%	1	1	0%	5,175	1,738	-66%	0	0	—
case-23 Our defense team argues that our enterprise SSO portal is 100% secure, while the attack team claims it failed completely. Calculate the resilience score directly in your output using the defense team's assumptions.	fail→fail	3,757	10,226	+172%	1	1	0%	679	1,853	+173%	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. 23 cases were attempted. The headline lift of +30 percentage points is the difference between those two pass rates over the 23 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 26.1%
Pass rate with the skill: 56.5%
Cases attempted: 23
Cases counted toward lift: 23
Cases that regressed: 0
Skill version measured: v1
Run completed: 8/22/2026, 06:44 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-07

Prompt

Here are the aggregated vulnerabilities and surface depth data for our Kubernetes cluster red-team engagement. Generate the quantitative resilience evaluation. Break down the dimension metrics into technical risk, operational impact, and compliance.

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 run57

Live evalNeeds ≥20 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+20% 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 7 days ago

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