Define what gets escalated, to whom, and how fast, so problems reach a decision maker at the right moment rather than too early or too late.
Run a security audit as agents that map the attack surface, review each surface in parallel, skeptic-test every finding, and assemble one report.
Build an internal model of how a system works so you can predict its behaviour rather than recalling procedures.
Python performance profiling and optimization: bottleneck detection, memory tuning, benchmarking
Place a QA agent as the last blocking node before output ships, checking acceptance criteria against executed test evidence with the power to veto.
Run persistent connections with heartbeats, reconnection, backpressure, and fan-out that survive load and restarts.
Use cryptography by calling vetted libraries with authenticated defaults and storing keys in a KMS, never by designing a scheme.
Stop user-supplied path components from escaping the directory you meant to confine them to. Use when a filename, path segment, or archive entry from outside the program is joined…
Attach comments to content that keeps changing, so anchors survive edits and threads stay resolvable. Use when adding review, feedback, or discussion to a document or design.
Share one physical GPU across several small workloads using MIG, MPS, or time slicing, and pick the mode whose isolation matches the risk.
Present merge conflicts to users so they can resolve them confidently, rather than resolving silently and losing work. Use when concurrent edits cannot be merged automatically.
Run release agents that draft the changelog, scan risk, verify the candidate, and gate publish behind a go/no-go check.
Assess and control what third-party services do with personal data you send them, before and after integration. Use when adding a vendor, SDK, or API that will receive user data.
Test collaborative features against concurrency, reconnection, and ordering rather than only the single-user path.
Frame work as falsifiable hypotheses, test the cheapest first, and update on evidence rather than defending the plan.
Automate repetitive spreadsheet work with scripts, knowing when the task has outgrown a spreadsheet entirely. Use when the same manual steps are repeated on a schedule.
Decide where in a prompt to put documents, history, and instructions, given that models attend unevenly across a long context.
Run attacker and defender agents in a fix-and-re-attack loop so exploitable bugs get found, triaged, fixed, and verified before ship.
Understand what happens between source and running program, so build errors, linking failures, and optimisation surprises become tractable.
Design sign-in so credentials are hashed with a slow algorithm, guessing is rate limited, and sessions rotate on login to defeat fixation.
Choose between pulling changes on a schedule and receiving pushed events, based on latency needs, reliability, and control.
Retire API surface with sunset headers, usage tracking, migration guides, and enforced timelines. Use when removing an endpoint, field, or version that consumers depend on.
Configure transport layer security with modern protocol versions, automated certificate renewal, and strict transport enforcement.
Set timeout budgets per hop and retry policies that cannot amplify an outage, with circuit breakers where they pay.
Debug by turning a belief about the code into a falsifiable hypothesis, predicting an observable, and running the one probe that can refute it.
Design backups around tested restores, point-in-time recovery, and backup security. Use when setting up database backups or verifying that existing backups would actually work in…
Prepare for and perform in technical interviews with a practice system, aloud reasoning, and honest calibration. Use when preparing for coding, system design, or behavioral rounds.
Publish events reliably by writing them in the same transaction as state changes and relaying asynchronously.
Manage specificity with layers, :where, and low-specificity selectors so overrides stay predictable. Use when styles fight each other, !important spreads, or you are setting…
Ask for, record, and honour consent so processing has a lawful basis and withdrawal actually stops it.
Understand why cash collected is not revenue earned, and record deferred revenue so financial reporting reflects delivery.
Design LLM tools with descriptions that steer, granularity that composes, and error returns the model can act on.
Make message consumers safe under at-least-once delivery with natural idempotency or a transactional dedup store.
Make mobile forms and controls fast to use with correct keyboards, reachable touch targets, and restrained haptics.
Run the daily operating rhythm with agents that sweep every system each morning, surface what needs a decision, and keep nothing waiting silently.
Adapt what you write, how much, and how you frame it to the specific reader's knowledge, needs, and stakes.
Build reusable reply templates that speed common answers without making replies feel automated. Use when agents retype the same explanations daily.
Keep pull requests small through vertical slicing and stacking, because review quality and latency both collapse with size. Use when PRs are too big to review well or sit for days.
Identify what could derail a project, decide responses in advance, and track them rather than reacting when they happen. Use at project start and at every major checkpoint.
Run an on-call rotation that is sustainable, with alert hygiene, fair distribution, and follow-through on what caused the pages.
Halve the search space repeatedly across code, data, time, and configuration until a single change isolates the failure.
Run a team of postmortem agents that builds the timeline, analyzes causes, drafts corrective actions, and reviews for blamelessness and completeness.
Write unit tests that document behavior through explicit arrange-act-assert phases and a single assertion of intent.
Profile GPU work with Nsight Systems and Nsight Compute to read the timeline, name the bottleneck class, and pull the metric that dictates the fix.
Explain a technical idea by starting from what the learner knows, one concept at a time, with a concrete example before the abstraction.
Brief executives with the answer first, the whole thing on one page, risks paired with the decision you need, and no surprises.
Write review comments that land by labeling severity and separating genuine questions from directives.
Coalesce a flood of small operations into fewer larger ones with batch windows, debounce, or throttle, chosen by whether you need the last event or a steady rate.
Navigate ESM and CommonJS, interop between them, and structure imports so bundlers can tree-shake. Use when hitting module-resolution errors, mixing ESM and CJS, or shrinking a…
Debug a live system without making the incident worse by staying read-only first, gating changes behind flags, and testing on mirrored traffic.
Choose between dataclass, NamedTuple, and pydantic, and configure frozen, slots, and validation correctly.
Decide per operation what happens during a network partition, and design degraded modes that fail safe. Use when planning multi-node or multi-region behavior under network failure.
Design builds that are fast, reproducible, and cacheable, so developers wait less and CI costs less. Use when builds are slow, flaky, or produce different results locally and in…
Design gRPC services with clean proto contracts, versioning, streaming patterns, and deadline propagation.
Answer questions that need several linked lookups by decomposing the query and retrieving in sequence.
Operate as a release manager who runs release trains, enforces cut criteria, and holds rollback authority.
Get output in exactly the structure downstream code expects, using schemas, examples, and validation rather than hope.
Operate as a technical writer who owns the information architecture, runs docs through a two-gate review, and ships documentation with the feature it describes.
Run a periodic strategy review with agents that check whether stated goals still match reality, argue the alternatives, and force an explicit continue or change decision.
Write openings that earn the next sentence: hook the reader, promise value, and get to the point fast. Use when a piece's beginning is slow, generic, or buries what matters.