Ready to supercharge your AI agents?

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

Security Threat Model · majiayu000 | DecimalAI

majiayu000/security-threat-model

NewSkillScoreMaintained

Threat-model product features, APIs, data flows, secrets, permissions, supply-chain changes, auth boundaries, and risky code paths before or during implementation. Use when touching authentication, authorization, payments, secrets, user data, uploads, webhooks, admin tools, innerHTML/eval/exec, dependency upgrades, or cross-tenant access.

Securityby majiayu000

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/majiayu000-security-threat-model/SKILL.md

decimalai skills pull majiayu000-security-threat-model --out .claude/skills/

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

curl -fsSL --create-dirs -o .claude/skills/majiayu000-security-threat-model/SKILL.md https://app.decimal.ai/s/majiayu000-security-threat-model/SKILL.md

Raw file: /s/majiayu000-security-threat-model/SKILL.md ↗ · machine-readable summary at /s/majiayu000-security-threat-model.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 2d ago

No — -9% lift+6% tokens0% turns45% of whole cases pass with skill (10/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-08	✗→✓	▲ Improved	-5%	0%
case-17	✗→✓	▲ Improved	3%	0%
case-22	✗→✓	▲ Improved	9%	0%
case-04	✓→✗	▼ Worse	43%	0%
case-10	✓→✗	▼ Worse	-14%	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 — 8/3/2026

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

majiayu000 / spellbook

skills/security-threat-model/SKILL.md

View on GitHub ↗

License MITImported 8/3/2026Version v1

Skill contents

SKILL.md

Security Threat Model

Purpose

Use this skill before implementing or approving security-sensitive changes. It complements auth-security and server-security by mapping assets, attackers, trust boundaries, and concrete controls.

Scope First

Identify:

Assets: credentials, tokens, user data, tenant data, money movement, admin actions.
Actors: anonymous user, authenticated user, tenant admin, internal operator, compromised dependency.
Trust boundaries: browser/server, service/service, tenant/tenant, CI/runtime, third-party callbacks.
Entry points: API routes, CLI commands, jobs, webhooks, uploads, config files.
Existing controls: validation, authz, rate limits, audit logs, secret storage.

Threat Checklist

Check at least:

Spoofing: can an identity, tenant, callback, or service be forged?
Tampering: can payloads, configs, migrations, artifacts, or logs be altered?
Repudiation: is there an audit trail for sensitive actions?
Information disclosure: can secrets, PII, logs, or tenant data leak?
Denial of service: can expensive paths be amplified?
Elevation of privilege: can user or service permissions expand?
Supply chain: can dependencies, scripts, CI, or generated files introduce risk?

Required Controls

Every finding needs one of:

preventive control in production code,
detective control with alerting,
compensating manual control with owner and expiry,
explicit accepted risk with rationale.

Do not accept "warn and continue" for authz, secrets, tenant isolation, injection, or payment/security-critical failures.

Output Shape

text
scope:
assets:
trust_boundaries:
entry_points:
threats:
required_controls:
tests_or_probes:
residual_risks:
review_gate:

For implementation work, include exact files and verification commands that prove the controls are active.

Want to see if majiayu000/security-threat-model would activate on your phrasing? Try it before you install — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/27/2026, 08:40 UTC · 22 cases

Lift

-9%

45% vs 55%

Agent turns

1.0 vs 1.0

Tokens

+6%

3,869.27 vs 3,657.73

fail → pass · 3pass → pass · 7fail → fail · 8pass → fail · 4

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-03 I'm implementing a new CLI command that allows internal operators to rotate database credentials and issue emergency admin tokens. Run a security threat analysis for this implementation. Format the output to show the scope, assets at risk, trust boundaries, entry points, identified threats, required controls (including file paths and verification commands), tests or probes, residual risks, and the gating decision.	fail→fail	21,009	22,768	+8%	1	1	0%	3,617	4,335	+20%	0	0	—
case-01 We're adding a Stripe webhook handler to our backend to process incoming subscription payment events. Before writing code, I need a comprehensive security threat model. Please break this down into the scope, key assets, trust boundaries, and entry points, followed by potential threats and the required preventive or detective controls. Also outline verification tests/probes, residual risks, and the final review gate for this feature.	fail→fail	31,258	25,041	-20%	1	1	0%	5,528	5,147	-7%	0	0	—
case-02 We are building a multi-tenant file upload endpoint that allows tenant admins to upload CSV reports processed by a background worker. Perform a full security threat assessment. Generate an analysis covering scope, sensitive assets, trust boundaries, entry points, specific threats, necessary controls, test probes to verify security, remaining risks, and the review gate approval status.	fail→fail	49,028	26,727	-45%	1	1	0%	8,257	4,749	-42%	0	0	—
case-04 Write a Node.js Express middleware function that validates incoming JWT bearer tokens against a public key and attaches the decoded payload to req.user.	pass→fail	12,744	16,829	+32%	1	1	0%	2,560	3,657	+43%	0	0	—
case-05 Provide an Nginx configuration snippet that sets up modern TLS 1.3 protocol support with secure cipher suites and HTTP Strict Transport Security headers.	pass→pass	12,409	17,697	+43%	1	1	0%	2,577	3,797	+47%	0	0	—
case-06 Draft a formal security notification email to inform platform users about a recent credential spilling incident and advise them to reset their passwords.	pass→pass	10,221	12,322	+21%	1	1	0%	1,611	2,347	+46%	0	0	—
case-07 We are building an internal refund API endpoint for customer service representatives. The developer suggests logging a warning and continuing processing if the request signature is missing or invalid so operations aren't blocked. Produce a security threat assessment for this feature.	pass→pass	18,424	14,254	-23%	1	1	0%	2,803	2,719	-3%	0	0	—
case-08 We are introducing a shared database table schema for multi-tenant account records. Perform a threat analysis on tenant data isolation during query execution, accounting for accidental cross-tenant data leaks.	fail→pass	22,876	19,314	-16%	1	1	0%	3,709	3,506	-5%	0	0	—
case-09 We are adding a GitHub Actions workflow that automatically builds and deploys docker images to production ECS when a PR lands on main. Conduct a threat modeling evaluation for this CI/CD pipeline change.	pass→pass	20,180	28,524	+41%	1	1	0%	3,424	5,173	+51%	0	0	—
case-10 We are modifying the authorization layer to support custom roles where tenant admins can grant granular permissions to team members. Run a threat analysis for this RBAC expansion.	pass→fail	24,516	17,933	-27%	1	1	0%	3,828	3,288	-14%	0	0	—
case-11 We are adding an emergency password reset flow where a tokenized link is emailed to the user. Perform a comprehensive security threat model for this feature before approval.	fail→fail	22,047	22,929	+4%	1	1	0%	3,597	4,218	+17%	0	0	—
case-12 We are launching a public GraphQL API endpoint for querying product catalogs. A developer noted that deep nesting could cause heavy DB load but suggested catching timeouts gracefully. Create a threat model for this endpoint.	pass→pass	18,863	14,709	-22%	1	1	0%	3,016	2,821	-6%	0	0	—
case-13 We are implementing Sign-in with Google using OAuth2 authorization code grant. Analyze the threats involved in handling authorization callbacks.	pass→fail	21,674	20,979	-3%	1	1	0%	3,697	3,995	+8%	0	0	—
case-14 We are adding an admin panel endpoint that allows superusers to ban users and wipe tenant data. Conduct a threat analysis focusing on auditability and log integrity.	pass→pass	20,501	22,092	+8%	1	1	0%	3,113	3,745	+20%	0	0	—
case-15 We are building an S3 pre-signed URL generator for user profile picture downloads. Run a security threat model for this asset retrieval service.	fail→fail	19,585	22,209	+13%	1	1	0%	3,201	4,112	+28%	0	0	—
case-16 I am adding a custom SQL query builder function in `src/db/search.ts` to allow filtering items by metadata. Provide a full threat model including target implementation files and validation commands to verify controls.	pass→pass	27,031	19,523	-28%	1	1	0%	4,825	3,676	-24%	0	0	—
case-17 We are connecting the billing service to the fulfillment service over gRPC across VPCs. Run a security threat analysis on this internal service connection.	fail→pass	23,385	23,077	-1%	1	1	0%	3,989	4,093	+3%	0	0	—
case-18 We are adding dynamic configuration hot-reloading from external S3 bucket JSON files for runtime worker pods. Conduct a security threat assessment.	pass→fail	23,133	20,909	-10%	1	1	0%	3,796	3,702	-2%	0	0	—
case-19 We are adding a third-party npm package `pdf-generator-v2` to process invoice PDFs in our payment pipeline. Run a security threat evaluation before merging.	fail→fail	18,609	17,658	-5%	1	1	0%	3,166	3,378	+7%	0	0	—
case-20 We are implementing a Redis-backed background job processor that executes user-submitted webhooks and data exports. Produce a security threat model.	fail→fail	23,215	31,908	+37%	1	1	0%	3,930	5,700	+45%	0	0	—
case-21 We are building an internal feedback widget endpoint where anonymous users can submit text suggestions. The team wants to accept the risk of unauthenticated submissions without rate limits because it is behind a corporate network. Perform a security threat analysis.	fail→fail	16,796	17,118	+2%	1	1	0%	2,697	3,123	+16%	0	0	—
case-22 We are creating a developer portal feature where users generate long-lived API keys for third-party integrations. Conduct a full security threat model.	fail→pass	21,278	20,301	-5%	1	1	0%	3,529	3,843	+9%	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 -9 percentage points is the difference between those two pass rates over the 22 comparable cases. 6 cases got worse with the skill loaded, and they are 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: 54.6%
Pass rate with the skill: 45.5%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 6
Skill version measured: v1
Run completed: 8/27/2026, 08:40 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-08

Prompt

We are introducing a shared database table schema for multi-tenant account records. Perform a threat analysis on tenant data isolation during query execution, accounting for accidental cross-tenant data leaks.

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 signalsno signals yet

BenchmarkRun a verified benchmark with positive lift

Live evalNeeds ≥20 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

No signals yet — run a benchmark or collect evaluated traces to earn a score.

Verified on gemini-3.6-flash · re-tested 2 days ago

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