Test design, vulnerability assessment, and quality assurance (13K skills)
Investigate WorldMonitor production errors using Sentry event details and source code. Classify issues as noise, already fixed, product bugs, ingestion problems, or needing human review, with tested fixes when code changes are requested.
Retrieve current cyber threat data, including malware indicators, addresses of attacker control servers, and vulnerabilities known to be exploited. Filter by threat type, source, severity, and date to identify relevant threats for review and reporting.
Evaluate slow, unreliable, duplicated, or brittle tests by the behavior they protect and their maintenance cost. Use it to choose focused checks, justify additions or removals, and distinguish product bugs from outdated expectations or environment failures.
Find Python input/output calls that may stall asynchronous backend work, then fix or guard each relevant path with runtime tests. Verify that tests fail when blocking behavior returns and pass after the fix is restored.
Check a DeerFlow installation from code updates through deployment and basic service checks. Supports local and Docker setups, verifies frontend pages and API health, and produces a report with results and troubleshooting recommendations.
Review existing DeerFlow skill packages for publication readiness, activation rules, safety boundaries, supporting files, and evaluation evidence. Receive concrete findings and suggested replacement text, with a clear distinction between inspection results and verified runtime behavior.
Develop Rust filters for RTK through tests written before implementation. Use real command output as test inputs, check output snapshots and unexpected inputs, verify a minimum reduction target, and complete formatting, linting, and test checks before finishing.
Guide Rust development through test-driven development: write a failing test, implement enough code to pass, then refactor. Includes practical test patterns and final checks for formatting, static analysis, and test results.
Run or inspect local Release-Green checks for a specific code revision or combined pull request candidate. Report blocking failures, diagnostic differences, and missing evidence, compare against the base revision, and identify remaining CI checks.
Create and run automated tests that score LLM answers against expected outputs or judging criteria. Monitor progress, inspect individual results, compare models on the same tests, and use score thresholds in CI, an automated checking workflow.
Run scoped quality checks without changing product code. Record the exact revision, commands, logs, and outcomes, then report failures, incomplete checks, and existing unresolved issues to support reviews or release preparation.
Review OmniRoute audit logs and request history from the command line. Filter records by actor, action, or time, manage access rules, and export logs or compare operational data for compliance reviews.
Test local web applications with Python scripts using Playwright, a browser automation tool. Verify interface behavior, investigate problems, capture screenshots, and inspect browser logs, with a helper that manages one or more application servers.
Check whether an application is ready for production by finding unfinished code and testing real database and API connections. Review performance under load, authentication, environment settings, and deployment behavior to identify gaps before release.
Benchmark distributed consensus protocols, methods computers use to agree on shared results. Compare processing rates, response times, scaling behavior, and resource usage across Byzantine, Raft, and Gossip, then identify bottlenecks and potential tuning changes.
Measure throughput, response times, and resource use for systems and groups of agents. Use baseline comparisons, load tests, and performance criteria to identify slowdowns, assess behavior under pressure, and summarize results with recommendations.
Coordinate agreement across distributed participants when some may fail or act maliciously. It covers PBFT, a three-phase agreement protocol, along with message authentication, malicious participant detection, leader replacement, and recovery from network partitions.
Scan code that handles sign-in, access permissions, payments, or user data for security weaknesses. The skill provides commands for a full security analysis and focused checks of input validation, which checks whether incoming values are acceptable.
Check a Claude Code plugin before publishing. Inspect file placement, required configuration fields, metadata at the start of skill and agent files, and MCP tool references, then report failures with practical suggestions for fixing them.
Check a codebase for known dependency vulnerabilities, hard-coded secrets, and risky code patterns using Ruflo security tools. Choose a scan depth, inspect matching lines, and report verified concerns with a clear account of what was checked.
Find code that lacks test coverage and get suggestions for tests to add next. Focus on a chosen path, inspect gaps against a coverage threshold, or use a background worker for recurring checks after feature changes.
Audit project dependencies for known vulnerabilities and outdated packages. Organize findings by severity, review proposed package changes before optional automatic fixes, and rerun the audit to check whether the reported vulnerability count decreased.
Review implemented features against their requirements, add missing tests, and check performance constraints. Produce documentation, a table linking requirements to code and tests, and evidence showing whether the feature is ready for deployment.
Investigate failed GAIA benchmark questions by inspecting execution records and classifying causes such as missing tools, reasoning errors, or answer extraction bugs. Follow targeted fixes, verify them through another run, and save findings for future debugging.