Development
Write code that keeps untrusted input, secrets and privileges in their place, with dependencies, reviews and a build pipeline you can trust.
Does the code keep untrusted input, secrets and privileges where they belong, and can the pipeline that builds it be trusted?
Start here: Fix a finding
Security practices in this phase
Secure coding
Use the safe API for each sink: bound parameters for SQL, argument arrays for commands, context-aware encoding for HTML, and allowlists for redirects and outbound requests.
Done when: Each weakness class the change touches has a test that fails on the vulnerable pattern.
Secrets out of code
Load credentials at runtime from a secrets manager or the platform, and scan every commit for secrets.
Done when: No secret is in the repository or its history, and a leaked one can be rotated without changing code.
Dependency selection
Check a new package’s maintenance, provenance and install scripts before adding it, pin it with a lockfile, and install only from the registry you intend.
Done when: New dependencies are reviewed in the pull request, and the lockfile pins every one.
Security code review
Review pull requests for the boundaries they touch: input parsing, authorization checks, output encoding, secrets and error handling.
Done when: Branch rules require an approving review from someone other than the author before merge.
AI coding assistants
Review assistant output like any other contribution, keep secrets out of prompts and context, and limit what coding agents may run or change.
Done when: Generated code goes through the same review and checks, and agents run with scoped permissions.
CI security checks
Run static analysis, secret scanning and dependency checks on every pull request, and decide in advance which findings block a merge.
Done when: The blocking criteria are written down, findings are triaged within an agreed time, and exceptions expire.
Pipeline hardening
Pin actions to commit SHAs and images to digests, give jobs least-privilege tokens, keep untrusted pull-request code away from secrets, and protect release branches.
Done when: Workflow permissions are read-only by default, and release jobs cannot be triggered from forks.
Guides
121 guides- Fix SQL injection in Python sqlite3
Bind query values separately from SQL syntax, then prove the fix preserves legitimate behavior and blocks an injection probe.
- Enforce authorization for every object action
Authorize the authenticated subject, requested operation, and specific resource together; a valid login or unpredictable identifier is insufficient.
- Design a safer file-upload pipeline
Treat upload receipt, content processing, storage, and later delivery as separate security decisions.
- Verify webhook deliveries before acting
Verify the provider’s signature over the correct bytes, then apply replay, authorization, and idempotency controls before a business action.
- Protect the path from pull request to release
Separate untrusted contribution checks from privileged release work and verify the identity of the artifact that is actually deployed.
- Contain prompt injection at the application boundary
Assume retrieved content can influence model output; enforce data access and consequential actions in components the model cannot authorize for itself.
- Secrets management: storage, rotation and leaked-secret response
Follow one application credential from issuance through caching, rotation, revocation and restored backups, prove that its old value stops working, and respond correctly when a secret leaks.
- Review trust boundaries in mobile and native clients
Protect local data and platform interactions while keeping server-side authorization independent of the client’s behavior.
- ECB mode: why it leaks patterns and what to use instead
Derive exactly what ECB reveals and why a longer key cannot fix it, find where ECB enters real code, and replace it with an authenticated mode.
- Padding oracle attacks
Use a byte-level CBC example to see how a tiny validity signal becomes plaintext information, then design verification that checks more than the error message.
- RC4 keystream bias
Understand why fresh RC4 keys do not remove keystream bias and how repeated secret positions turn a small statistical effect into evidence.
- 64-bit block ciphers and the Sweet32 mechanism
Calculate the birthday effect for 64-bit blocks and derive the CBC relation that can turn a collision into a plaintext disclosure.
- Prevent XSS at the output boundary
Choose text rendering, context-aware encoding, or deliberate HTML sanitization according to how the browser will interpret the value.
- Constrain server-side fetches to their intended destinations
Prevent untrusted URLs from turning a preview, import, or webhook feature into access to internal services.
- Session management: cookies, timeouts, rotation and revocation
Model authentication as transitions in server authority, set session cookies and timeouts to current NIST and OWASP values, and verify fixation prevention, expiry, reauthentication, revocation and stolen-cookie defenses.
- Password storage: hashing parameters and rehash on login
Choose Argon2id, scrypt, PBKDF2 or bcrypt with current minimum parameters, store a versioned hash, and upgrade old hashes after a successful login without overwriting a reset.
- Bound API work, not just request counts
Protect availability and operating cost with limits on input, per-request work, concurrency, queues, and downstream spending.
- Handle failures without leaking internals or weakening controls
Give callers a useful, bounded response while preserving protected diagnostic evidence and safe application state.
- Ship without debug modes, verbose errors or exposed files
Turn off debug output, publish an allowlisted build and probe each deployment for exposed files and endpoints.
- Protect cookie-authenticated actions against CSRF
Make the server distinguish an intended action from a request another site caused the browser to send.
- Configure CORS and CSP around the browser capabilities you need
Control who can read API responses and which resources a page may execute without confusing either control with authorization.
- Render rich text without granting it application privileges
Choose text rendering or a restricted HTML policy, then test the final DOM after every transformation.
- Authorize functions and fields as well as objects
Prevent mass assignment, sensitive-field exposure and administrative actions that a valid object lookup alone cannot protect.
- Keep untrusted values out of executable syntax
Choose a safe interface for each interpreter and bound the work performed by parsers and converters.
- Run external programs without command or argument injection
Run programs from an argument list with no shell, end options before user values, and keep untrusted arguments away from Windows batch files.
- Keep untrusted input out of eval, expression languages and templates
Keep code and templates constant, pass input as data, and treat in-process sandboxes as one layer, not a boundary.
- Extract archives without escaping the destination
Prevent traversal, link tricks, overwrites and decompression exhaustion while preserving legitimate file extraction.
- Convert untrusted office documents and PDFs in an isolated parser
DOCX, XLSX and ODF files are ZIP archives of XML, and PDFs can carry scripts; the converters that read them are large C and C++ programs with a long CVE history. Check the package before any parser reads it, convert each document in a container with no network, no writable root and hard limits, kill it at a deadline, and re-render to pixels when people will open the result.
- Secure WebRTC signaling, TURN relays and media servers
WebRTC encrypts media, but signaling decides whose keys you trust and TURN relays wherever it is allowed. Authorize every signaling message, mint short-lived TURN credentials, deny internal addresses and harden media servers.
- Review a code change through its trust boundaries
Turn a pull request into a small set of security claims that can be checked against the surrounding system and tests.
- Preserve framework security controls across the request lifecycle
Trace middleware, persistence and rendering so custom code does not bypass the protections supplied by the framework.
- Review mobile storage, links and platform permissions
Keep local artifacts and platform entry points from exposing authority that belongs on the server.
- Constrain desktop IPC and privileged helpers
Keep renderer content, local callers and plugins from turning a narrow feature into unrestricted operating-system access.
- Design firmware updates that can recover safely
Connect trusted boot, update authorization, rollback policy and end-of-life behavior to the device’s actual failure modes.
- Maintain an API inventory and prevent gateway bypass
Connect documented routes, deployed listeners and backend trust so retired or direct paths do not bypass the intended controls.
- Apply security limits to GraphQL, WebSocket and gRPC operations
Carry authorization and resource budgets past the initial HTTP request into resolvers, messages and long-lived streams.
- Triage new dependency vulnerability findings
Turn a weekly batch of Dependabot, OSV-Scanner or npm audit findings into a short list of distinct, shipped, reachable issues, each routed to one outcome: fix, not affected, accept or escalate.
- Keep untrusted CI work outside the release boundary
Review workflow authority, persistent state and artifact handoffs before contributor-controlled code reaches privileged jobs.
- Investigate a suspected malicious package safely
Distinguish package substitution, publisher compromise and dangerous installation behavior while containing exposure and preserving evidence.
- Review dependency health without treating a score as a guarantee
Combine security practices, maintenance, licensing and product exposure into a recorded adoption or replacement decision.
- OAuth 2.0 and OpenID Connect login: authorization code with PKCE
Implement OAuth 2.0 and OpenID Connect login with the authorization code flow and PKCE, keep each transaction bound to its browser, and block redirect, mix-up, code-injection and token-theft attacks.
- JWT validation: algorithms, keys, claims and key rollover
Validate a JWT against a fixed algorithm, trusted keys and the expected issuer, audience and type, reject the known header attacks, and rotate signing keys without accepting arbitrary keys or breaking valid tokens.
- Passkeys and MFA: WebAuthn enrollment, sign-in and replacement
Register, verify and replace passkeys with WebAuthn so phishing resistance survives the account’s whole lifecycle, and map passkeys and other MFA methods to the NIST SP 800-63B-4 assurance levels.
- Account recovery: password resets, help-desk resets and account linking
Treat password resets, support-desk recovery and account linking as new grants of authority: single-use high-entropy proofs, atomic consumption, session revocation, and stable identity rather than email matching.
- AEAD records: AES-GCM nonces, associated data and key limits
Build an encrypted record whose tag binds ciphertext to the intended tenant, record, purpose and format, keep nonces unique within each key’s usage limit, and verify the design with actual AES-GCM.
- Workload identity federation: short-lived, scoped identities for services and CI
Trace how a running workload obtains a short-lived identity, write cloud trust policies that match immutable claims instead of names or wildcards, and keep the resulting credential limited by audience and authorization.
- Authorization models: RBAC, ABAC, ReBAC and policy as code you can test
Translate roles, attributes and relationships into explicit subject-action-resource decisions, prevent broken object level authorization (BOLA), and verify ownership and tenant boundaries with tests.
- Keep retrieval inside the user’s data boundary
Carry document permissions through ingestion, retrieval, caching and citations so a model never receives another tenant’s records.
- Authorize agent tools outside the model
Bind each proposed action to the user, resource and approved arguments before allowing a tool to cause a side effect.
- Keep generated output out of executable contexts
Choose safe rendering and narrow operations so model text cannot become browser code, database syntax or an unrestricted command.
- Bound agent spending, memory and recovery
Enforce budgets and stop conditions outside the model, and make retained context and side effects recoverable.
- Treat model releases as executable supply-chain inputs
Verify provenance, constrain loading and preserve the relationship between model weights, configuration and runtime dependencies.
- Protect training and retrieval data from targeted poisoning
Track who can influence datasets, keep evaluation independent and investigate targeted behavior changes without relying on aggregate accuracy.
- Assess extraction, membership and inference risks
Define what an adversary can observe and measure privacy leakage without confusing a model’s output with proof about its training data.
- Constrain MCP and multi-agent trust boundaries
Validate each connection and delegated action without treating a token for another service, a tool description, another agent or a state handle as authority.
- Contain cascading failures and rogue agents
Stop a looping, compromised or drifting agent from spreading harm: kill switches, budget and rate limits, circuit breakers, behavior monitoring, blast-radius limits, quarantine, rollback and incident response, all enforced outside the agent.
- Test smart contracts through state invariants
Define the properties funds and privileges must preserve, then exercise callbacks, transaction sequences and exceptional conditions locally.
- Parse bounded JSON and XML into an explicit application schema
Separate parser capabilities from schema rules, reject ambiguous and coercible inputs, and implement the boundary in a complete Python project.
- SAML single sign-on: verify signed assertions without losing the identity boundary
Bind a validated SAML assertion to its trusted issuer, the exact signed element, your service, the browser request and a replay record before creating a local session, and keep your SAML library ahead of its advisories.
- SCIM deprovisioning: make account disablement remove every usable authority
Connect SCIM resource updates to current sessions, tokens, group-derived permissions and background jobs, with explicit tenant mapping, CAEP revocation events and a measurable revocation contract.
- Keep untrusted object keys out of prototype and authority paths
Trace an actual nested-merge mutation, replace it with a narrow schema, and make authorization independent of inherited properties.
- Harden an Android app's storage, network and entry points
Encrypt local secrets with Android Keystore keys now that Jetpack Security is deprecated, set transport policy in Network Security Configuration, close exported components and mutable PendingIntents, confine WebViews, verify App Links and judge Play Integrity verdicts on the server, mapped to OWASP MASVS 2.1.
- Prevent CRLF header injection, log forging and CSV formula injection
Carriage return and line feed separate HTTP header fields, log records and CSV rows, and a leading = turns a spreadsheet cell into a formula. Let the framework write headers, log structured JSON, and neutralize formula triggers in exports.
- Prevent DOM XSS, postMessage abuse and client-side injection
Trace attacker-controlled values from the URL, messages and storage to the browser APIs that parse HTML or run code, then close those sinks with safe APIs, a sanitizer, Trusted Types and strict message checks.
- Harden an iOS app's storage, network and entry points
Store secrets in the Keychain with the narrowest accessibility class, set file protection explicitly, keep App Transport Security on, prefer universal links, constrain WKWebView bridges, keep step-up keys in the Secure Enclave, verify App Attest on the server and ship an accurate privacy manifest, mapped to OWASP MASVS 2.1.
- Find, fix and contain memory-safety bugs in C and C++
Recognize out-of-bounds access, use after free, NULL dereference and integer overflow in C and C++, fix them with checked lengths and single ownership, harden release builds, catch the rest with sanitizers and fuzzing, and plan new code in a memory-safe language.
- Prevent NoSQL, LDAP, XPath and expression-language injection
SQL is not the only query language that receives user input. Stop MongoDB operator injection with type checks and $eq, escape LDAP filter values, pass XPath values as variables, and never evaluate SpEL or OGNL built from request data.
- Redirect users only to destinations you control
Check every redirect target taken from a request, so attackers cannot borrow your domain for phishing, carry OAuth tokens off-site or steer a server-side fetcher through your redirects.
- Validate input safely and prevent regex denial of service
Find and fix exponential and quadratic regex backtracking with measured examples, choose linear-time engines or timeouts, and build allowlist validation that decodes once and rejects type confusion and parameter pollution.
- Audit unsafe Rust, FFI boundaries and crate dependencies
Know what the Rust compiler proves and what it leaves to you: review unsafe code for soundness, wrap foreign calls in small safe functions, decide how overflow and panics behave in release builds, gate crates with RustSec, and replace C one boundary at a time.
- Configure HTTP security headers from a tested baseline
Send the response headers that make browsers enforce HTTPS, run only your own scripts, refuse hostile framing and keep cookies on one host, then prove they appear on every kind of response.
- Stop dependency confusion and typosquatting at install time
Pin every internal package name to its real source, verify installed bytes against a lockfile, and delay or gate install scripts so a public look-alike cannot replace a private dependency.
- Harden GitHub Actions against workflow and supply-chain attacks
Close the paths that give untrusted pull-request code secrets or a write token: pass context through env, keep pull_request_target away from fork code, pin actions to commit SHAs, scope GITHUB_TOKEN and OIDC, and treat artifacts as data.
- Protect the source repository: rulesets, reviews and signed commits
Configure the controls every repository that ships software needs: protected branches and tags, required review with CODEOWNERS, signed commits, secret push protection and account security, mapped to the SLSA Source track, the OpenSSF OSPS Baseline and Scorecard.
- Secure the developer workstation: extensions, package managers, secrets and AI agents
Attackers reach the software supply chain through developers' editors, package installs and AI agents. Allow-list extensions, stop install scripts and brand-new versions, keep long-lived secrets off laptops, and lock down repository files that run code when a project opens.
- Open source license compliance: SPDX expressions, obligations and SBOM checks
Know which licenses you ship, read SPDX license expressions correctly, meet the obligations of common licenses for the way you distribute, and enforce a license policy in CI from the SBOM.
- Meet the OpenSSF OSPS Baseline as a maintainer: the levels, every Level 1 control on GitHub, and the evidence
Walk through the eight control families and three maturity levels of OSPS Baseline v2026.08.28, meet each Level 1 requirement on a GitHub repository, and keep evidence a consumer can check.
- Adopt the OpenSSF S2C2F as a consumer: the eight practices, the four maturity levels and a staged plan with evidence
Assess how your organization takes in open source against the Secure Supply Chain Consumption Framework's 25 requirements, then reach each maturity level in turn with evidence an assessor can check.
- Choose cryptographic algorithms and parameters by job
A sourced quick reference for October 2026: which algorithm and parameters to use for encryption, hashing, passwords, MACs, signatures, key agreement, key derivation, randomness and TLS, the matching APIs in Node.js, Python, Java, Go and .NET, and where post-quantum migration stands.
- Defend JWTs against forgery, confusion and replay
Verify a JWT as an issuer-, type- and audience-bound assertion with an algorithm allowlist and trusted keys, closing the alg "none", key-confusion, header-injection, weak-secret and replay attacks.
- Contain a leaked secret, then investigate and clean up
Revoke an exposed API key, token or private key at its issuer before anything else, end the sessions it minted, check the provider's logs, then remove every copy and replace it with a credential that expires on its own.
- Stop account takeover with the right MFA and login defenses
Rank MFA options by the attacks they stop, implement TOTP and push approvals correctly, and close the login, reset and recovery paths that attackers use instead of breaking the second factor.
- Defend OAuth and OIDC against redirect and token attacks
Apply RFC 9700 and RFC 10017: exact redirect matching, PKCE with downgrade protection, state and iss binding, refresh-token rotation, sender-constrained tokens, and a backend-for-frontend so browser tokens cannot be stolen.
- Secure AI-assisted development with coding assistants and agents
Treat generated code, suggested packages, repository text and MCP servers as untrusted inputs, and confine coding agents with OS-level isolation and short-lived, narrowly scoped credentials.
- Secrets at runtime: Vault, Kubernetes Secrets, CSI driver, External Secrets and SOPS
How a running workload should get its credentials: encrypt Kubernetes Secrets with a KMS v2 provider and lock down RBAC, mount from an external store with the Secrets Store CSI Driver, sync with External Secrets Operator, issue dynamic credentials with Vault and its Agent Injector, keep Git copies encrypted with SOPS, and make rotation reach the process.
- Prevent file inclusion: LFI, RFI, dynamic imports and template names
Stop request data from choosing which file, module or template your application loads: local and remote file inclusion in PHP, dynamic require() and import() in Node.js, importlib and template names in Python, Thymeleaf view names in Spring, render in Rails, and the allowlist fix that works in all of them.
- Create temporary files safely: mkstemp, O_EXCL and private directories
Stop symlink, hardlink and pre-created-file attacks on the temporary files your code writes in shared directories such as /tmp: create each file atomically with O_EXCL and owner-only permissions through mkstemp or its equivalent in shell, Python, Java, Node.js, Go, .NET and Rust, keep related files in a private directory, and treat the Linux protected_* sysctls and systemd PrivateTmp as a backstop.
- Consume third-party APIs safely (OWASP API10:2023)
Treat every upstream and third-party API response as untrusted input: validate it against a schema, set connect and read timeouts, cap the response size, refuse to follow redirects blindly, keep TLS verification on, and scope the credentials you hold for each provider.
- Contain LLM misinformation (OWASP LLM07:2026)
Stop plausible but wrong model output from driving decisions and actions: ground answers in sources, verify every citation, separate generation from execution with a claim-check-act pattern, gate consequential actions on a human, and evaluate factuality on your own domain.
- Call a hosted LLM API safely
A code-first checklist for wiring up an OpenAI or Anthropic API call from a backend: keep keys server-side, redact PII before prompts and logs, choose provider retention settings deliberately, validate structured output and still treat it as untrusted, and set timeouts, retries and per-user cost caps.
- DeFi attack patterns: oracles, flash loans, proxies and signatures
The four smart-contract attack classes beyond reentrancy: price-oracle manipulation, flash-loan amplification, upgradeable-proxy storage collisions and uninitialized implementations, and signature replay, each with vulnerable and fixed Solidity, Foundry tests and verified incidents.
- Sandbox model-generated code and computer-use agents
Run the code an agent writes, and agents that drive a browser or desktop, inside a boundary built for hostile code: a disposable runtime with no secrets and no default network, hard resource limits, and a person in front of every side effect.
- Keep secrets and controls out of hidden context (OWASP LLM08:2026)
OWASP LLM08:2026 Hidden Context Exposure, which replaced System Prompt Leakage: why system prompts, tool schemas and retrieved policy text get extracted, what must never be in them, and how to enforce rules where a leak changes nothing.
- Defend agent memory and context against poisoning (OWASP ASI06)
OWASP ASI06 Memory & Context Poisoning: how an instruction or false fact saved once in agent memory, conversation summaries or a retrieval store steers later sessions, users and agents, and the write, scope, provenance, expiry and recovery controls that contain it.
Also part of this phase
- WEP: RC4 key reuse and weak IVs
Separate WEP’s repeated-keystream failure, weak-IV key recovery, and unkeyed integrity weakness through a concrete two-packet exercise.
- Lucky Thirteen and POODLE: two different oracles
Compare two CBC oracles by their signal, protocol rules, and attacker prerequisites instead of treating them as interchangeable padding bugs.
- DROWN: one RSA key across incompatible protocols
Map an RSA key across services to see how an SSLv2 oracle could compromise TLS traffic on a different endpoint.
- Sweet32: the cost of a small block size
Read the 2016 Sweet32 demonstrations as evidence about traffic volume, repeated secrets, and session lifetime rather than treating a block-size threshold as an instant exploit.
- Run Kubernetes workloads with only the privileges they need
Review pod settings, service accounts and effective permissions together before treating a namespace as an isolation boundary.
- Prove network and admission policies enforce the intended boundary
Verify allowed traffic and rejected deployments in the cluster instead of treating policy files as evidence of isolation.
- Bound background jobs and make retries safe
Keep queues, workers and paid dependencies within a defined budget when clients retry, cancel or disappear.
- Case study: the Shai-Hulud npm worm (2025-2026)
How a self-replicating npm worm turned stolen tokens into hundreds of infected packages in 2025, how its 2026 successors hijacked release pipelines and trusted publishing, what npm and GitHub changed in response, and the publishing, install and response practices that stop it.
- Case study: the tj-actions/changed-files compromise (CVE-2025-30066)
In March 2025 an attacker moved the version tags of a popular GitHub Action, and of a reviewdog action before it, to code that printed CI secrets into build logs. This case study traces the token chain back to one pull_request_target workflow, sizes the exposure, and shows how it was caught and which controls held.
- Case study: the xz Utils backdoor (CVE-2024-3094)
How a multi-year maintainer takeover planted a backdoor in xz Utils release tarballs, how build-time injection hid it from the git repository, why a performance anomaly exposed it, which systems shipped it, and what it teaches about reproducible builds and maintainer health.
- Case study: the Codecov Bash Uploader compromise (2021)
How a credential leaked through Codecov's Docker image process let an attacker change the Bash Uploader that many CI pipelines downloaded and ran, which sent their environment variables, and the secrets in them, to the attacker for two months in 2021.
- Case study: event-stream and ua-parser-js, two npm package takeovers
Two ways attackers took over popular npm packages: event-stream (2018), handed to a stranger who added a targeted, encrypted payload through a new dependency, and ua-parser-js (2021), whose maintainer's account was hijacked to publish versions that installed malware for four hours. Timelines, detection and the controls that stop both.
- Case study: the polyfill.io CDN takeover (2024)
How the sale of the polyfill.io domain and project in February 2024 turned a script tag on more than 100,000 websites into a channel for malicious redirects by June, why Subresource Integrity could not have protected it, and the controls that would have.
- Harden managed Kubernetes on EKS, AKS and GKE
Know what your provider secures and what you still own, then lock down API access, cloud identity, nodes, pod privileges and admission with settings that are generally available in Kubernetes 1.37, checked against the CIS and NSA/CISA baselines.
- Keep HTTP services up under application-layer DDoS
Bound every phase of an HTTP request, patch for the HTTP/2 attack classes, rate limit at the edge and the origin, and shed load deliberately, with nginx, Envoy and Node.js settings tested against slow clients and bursts.
- Harden container images from Dockerfile to admission
Build minimal, non-root images pinned by digest, keep secrets out of every layer, then scan, sign and verify the exact digest before it runs with dropped privileges.
- Harden the images you ship against CIS, STIG and SCAP benchmarks
Hold each image your team ships (container base images, golden VM images, Kubernetes node images) to a published benchmark through a pinned SCAP Security Guide profile, scan every build with OpenSCAP, and fail the build on any failure that a deviation register does not explain with an owner, an approver and an expiry.
- Case study: the LastPass breach (2022)
How an attacker went from a compromised developer laptop to a DevOps engineer's home computer, the keys to LastPass's cloud backups and a copy of customer vaults; why each vault's safety then rested on its master password and PBKDF2 iteration count; and what it teaches about backup keys, privileged endpoints, KDF upgrades and unencrypted metadata.
- Case study: the Okta support-system breach and session tokens in HAR files (2023)
How a service account password saved to a personal browser profile let an attacker download customers' HAR files from Okta's support system, replay the session tokens inside them against five customers, and stay unseen for two weeks; and why HAR files are credentials, admin sessions need binding, and every exposed credential must be rotated.
- Case study: Moffatt v. Air Canada, a chatbot that misstated the refund policy (2022-2024)
How an airline's website chatbot told a grieving customer they could claim a bereavement fare after travel, what the British Columbia Civil Resolution Tribunal decided in February 2024, and the grounding, testing and correction practices that keep an automated answer consistent with the policy it describes.
- Case study: the Nx s1ngularity compromise and its AI-assisted payload (2025)
How a pull-request-title injection in a GitHub Actions workflow on an outdated branch gave an attacker the Nx npm token in August 2025, how the malicious versions tried to turn developers' own AI command-line tools into file-search agents, and the workflow, publishing and agent controls that stop the same chain.
- Case study: the Amazon Q Developer extension compromise (2025)
How an over-scoped GitHub token, extracted from an AWS CodeBuild pull request build, let a threat actor commit code to the Amazon Q Developer extension for VS Code that shipped in version 1.84.0 in July 2025, why the injected code aimed at the Q Developer CLI failed, and the CI and AI-agent controls the incident argues for.
- Case study: Storm-0558 and the forged Microsoft tokens (2023)
How a stolen 2016 consumer signing key and a token validation gap let Storm-0558 read enterprise Exchange Online mail in 2023, and what the review board found about rotation, logging and culture.
- Case study: ingress-nginx CVE-2025-1974 and the exposed admission webhook (2025)
How an admission webhook open to the whole Pod network made ingress-nginx's configuration-injection bugs a 9.8 cluster-takeover risk in 2025, and what its 2026 retirement means.
- Case study: First American's EaglePro document exposure (2019)
How sequential IDs in unauthenticated links exposed First American's title documents in 2019, after its own penetration test found the flaw and a clerical error let it go unfixed.
- Case study: Spring4Shell (CVE-2022-22965)
How Spring's data binding let request parameters reach the class loader on JDK 9 and later in 2022, why the denylist workaround leaked, and what it teaches about framework-level mass assignment.
- Case study: CrowdStrike Channel File 291 (July 2024)
How a CrowdStrike content update crashed Windows hosts in July 2024: a field-count mismatch hidden by wildcard tests, a missing bounds check, and no staged rollout.