all publishers

SoluDevTech

@soludevtech source repo

68 published skills

  1. ▌
    Langchain RAG 2 · soludevtech
    INVOKE THIS SKILL when building ANY retrieval-augmented generation (RAG) system. Covers document loaders, RecursiveCharacterTextSplitter, embeddings (OpenAI), and vector stores (Chroma, FAISS, Pinecone).
    3 repo stars
  2. ▌
    Langchain Middleware 2 · soludevtech
    INVOKE THIS SKILL when you need human-in-the-loop approval, custom middleware, or structured output. Covers HumanInTheLoopMiddleware for human approval of dangerous tool calls, creating custom middleware with hooks, Command resume patterns, and structured output with Pydantic/Zod.
    3 repo stars
  3. ▌
    Langchain Dependencies 2 · soludevtech
    INVOKE THIS SKILL when setting up a new project or when asked about package versions, installation, or dependency management for LangChain, LangGraph, LangSmith, or Deep Agents. Covers required packages, minimum versions, environment requirements, versioning best practices, and common community tool packages for both Python and TypeScript.
    3 repo stars
  4. ▌
    Langchain Fundamentals 2 · soludevtech
    Create LangChain agents with create_agent, define tools, and use middleware for human-in-the-loop and error handling.
    3 repo stars
  5. ▌
    Langchain Python Quickstart 2 · soludevtech
    Scaffold a minimal local LangChain agent in Python by following the official quickstart. Use when the user wants to quickly build or try a LangChain agent locally.
    3 repo stars
  6. ▌
    Langchain Typescript Quickstart 2 · soludevtech
    Scaffold a minimal local LangChain agent in TypeScript by following the official quickstart. Use when the user wants to quickly build or try a LangChain agent locally.
    3 repo stars
  7. ▌
    Langchain RAG · soludevtech
    INVOKE THIS SKILL when building ANY retrieval-augmented generation (RAG) system. Covers document loaders, RecursiveCharacterTextSplitter, embeddings (OpenAI), and vector stores (Chroma, FAISS, Pinecone).
    3 repo stars
  8. ▌
    Langchain Middleware · soludevtech
    INVOKE THIS SKILL when you need human-in-the-loop approval, custom middleware, or structured output. Covers HumanInTheLoopMiddleware for human approval of dangerous tool calls, creating custom middleware with hooks, Command resume patterns, and structured output with Pydantic/Zod.
    3 repo stars
  9. ▌
    Langchain Dependencies · soludevtech
    INVOKE THIS SKILL when setting up a new project or when asked about package versions, installation, or dependency management for LangChain, LangGraph, LangSmith, or Deep Agents. Covers required packages, minimum versions, environment requirements, versioning best practices, and common community tool packages for both Python and TypeScript.
    3 repo stars
  10. ▌
    Langchain Fundamentals · soludevtech
    Create LangChain agents with create_agent, define tools, and use middleware for human-in-the-loop and error handling.
    3 repo stars
  11. ▌
    Langchain Python Quickstart · soludevtech
    Scaffold a minimal local LangChain agent in Python by following the official quickstart. Use when the user wants to quickly build or try a LangChain agent locally.
    3 repo stars
  12. ▌
    Langchain Typescript Quickstart · soludevtech
    Scaffold a minimal local LangChain agent in TypeScript by following the official quickstart. Use when the user wants to quickly build or try a LangChain agent locally.
    3 repo stars
  13. ▌
    Agno · soludevtech bundle
    Agno AI agent framework - build production-ready agents, multi-agent teams, workflows, MCP integrations, and deploy with AgentOS. Use when building, debugging, or learning about Agno agents.
    3 repo stars
  14. ▌
    Linter · soludevtech
    Run linting and formatting checks on the codebase. Use this skill when code changes are made and need static analysis validation. Runs ruff (Python) and eslint+prettier (TypeScript/React) depending on the project type. Invoke after implementation and before code review.
    3 repo stars
  15. ▌
    Githubpr · soludevtech bundle
    Manage the full GitHub PR lifecycle from a Jira ticket create a branch with the format <JIRA-ID>/<simple-description>, push, open a PR for review, poll CI, and merge when green. Use this skill whenever the user mentions creating a PR, opening a pull request, pushing a branch for review, or linking a Jira ticket to a GitHub PR. Also trigger when the user asks to wait for CI, merge a PR or manage the git workflow around a Jira ticket.
    3 repo stars
  16. ▌
    Sonarfix · soludevtech
    SonarQube issue remediation workflow. Use this skill when the user asks to fix SonarQube issues, run a code quality sweep, or remediate static analysis findings for a project. Retrieves issues via the SonarQube REST API using curl, groups them, presents a fix plan, and implements fixes in batches with test verification.
    3 repo stars
  17. ▌
    Trivyfix · soludevtech
    Trivy vulnerability remediation workflow. Use this skill when the user asks to fix Trivy vulnerabilities, run a security scan, or remediate CVE findings for a project. Runs trivy CLI scans, groups findings by severity and type, presents a fix plan, and implements fixes in batches with verification.
    3 repo stars
  18. ▌
    Dbanalyze · soludevtech bundle
    Analyze a database schema by running SchemaCrawler lint and extracting the full schema via information_schema. Use this skill whenever the user wants to audit a database schema, check normalization, detect missing indexes, redundant indexes, nullable FK columns, inconsistent data types, design smells, or get improvement recommendations.
    3 repo stars
  19. ▌
    Tester QA · soludevtech
    Manually test the app after a functionality is done. Two modes — Bug Hunt (find real bugs and persist them to <LOOP_DIR>/bug-reports/<slug>.md) and QA (write/fix Playwright specs in @soludev-compose-apps e2e repo). Use after the developer finishes writing code, tests, and documentation.
    3 repo stars
  20. ▌
    Iac Review · soludevtech bundle
    Static review of Kubernetes/Flux manifests in a GitOps repo (K3s + Flux CD). Scores manifests on 10 across correctness, security, reliability, consistency, GitOps hygiene, and maintainability. Use when reviewing PRs that change deployments, services, ingresses, network policies, external secrets, Flux Kustomizations, Helm values, or any K8s manifest — before it reaches the cluster. Complements popeyescan (runtime audit).
    3 repo stars
  21. ▌
    Popeyescan · soludevtech
    Run a Popeye scan on a Kubernetes/K3s cluster and produce a prioritized analysis. Use this skill whenever the user wants to audit cluster health with Popeye, run popeye, analyze popeye output, triage cluster warnings, check missing probes or resource limits, orphaned services, or review cluster quality score. Also use when the user pastes raw popeye output and asks for analysis, prioritization, or next steps.
    3 repo stars
  22. ▌
    Book To Skill · soludevtech bundle
    Converts books and documents (PDF, EPUB, DOCX, HTML, Markdown, plain text, RTF, MOBI/AZW with Calibre) into structured agent skills, extracting frameworks, mental models, principles, techniques, and anti-patterns. Use when the user wants to study a document through GitHub Copilot CLI, Amp, or Claude Code, apply an author's frameworks while working, or build a reusable knowledge base from a file.
    3 repo stars
  23. ▌
    Brainstorming · soludevtech bundle
    You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
    3 repo stars
  24. ▌
    Code Reviewer · soludevtech bundle
    Use to review the code, Invoke when you finished the task asked by the user. Grades the code on a scale from 1 to 10 across 6 dimensions (correctness, security, performance, maintainability, testability, architecture) with a per-dimension breakdown and an overall score that gates merging.
    3 repo stars
  25. ▌
    Code Simplifier · soludevtech
    Simplifies and refines code for clarity, consistency, and maintainability while preserving all functionality. Focuses on recently modified code unless instructed otherwise.
    3 repo stars
  26. ▌
    Performance Audit · soludevtech bundle
    Deep performance audit for Python and React codebases. Use this skill ANY TIME the user shares code and mentions performance, slow queries, latency, re-renders, optimization, profiling, bottlenecks, or memory issues — even if they don't use the word "audit". Also trigger when the user pastes code and asks "is this efficient?", "why is this slow?", "how can I improve this?", or shares ORM/database code (Django, SQLAlchemy, Prisma, TypeORM) where N+1 query risks are high. Covers: N+1 queries, missing indexes, loop anti-patterns, React re-render storms, missing memoization, bundle bloat, async misuse, memory leaks, caching strategy opportunities (query caching, HTTP caching, client-side caching, cache invalidation triggers), and 20+ other patterns across both stacks.
    3 repo stars
  27. ▌
    Test Writer React · soludevtech bundle
    Write unit and integration tests for React/TypeScript apps with Vitest, React Testing Library, MSW. Golden rule: real implementations for internal components (stores, hooks, providers), mocks only for outbound external adapters (APIs, SDKs), testcontainers for real-backend E2E. Trigger on: vitest, RTL, "write tests", "add tests", "test the component/hook/service", or any testing task in a React/TypeScript app. Use when testing any React app.
    3 repo stars
  28. ▌
    Linkedin Humanizer · soludevtech bundle
    Scrub AI tells from any text draft OR audit a finished post against the 2026 algorithm heuristic checklist. Tier-based rewriter (forensic / strict / aesthetic / all) plus `--mode audit` for detection-only pass-fail review covering length, hook, CTA, format penalties, AI vocab. Sub-tools: emoji-pattern detector, multi-detector spread tester (GPTZero, Originality.ai, ZeroGPT, Sapling, Copyleaks), rule explainer. Triggers on "humanize", "de-AI", "review this draft", "audit before posting", "is this ready".
    3 repo stars
  29. ▌
    Test Writer Nestjs · soludevtech bundle
    Write unit and integration tests for NestJS hexagonal backends. Jest, ts-jest, Supertest. Golden rule: real implementations for internal components (real TypeORM + SQLite in-memory), mocks only for outbound external adapters (email, Stripe, S3). Use when testing a use case, controller, service, or adapter in a NestJS app.
    3 repo stars
  30. ▌
    Test Writer Python · soludevtech bundle
    Write unit and integration tests for Python/FastAPI apps with pytest + pytest-asyncio. Golden rule: real implementations for internal components, mocks only for outbound external adapters (email, Stripe, S3), testcontainers for real Postgres/Redis/Kafka. Trigger on: pytest, "write tests", "add tests", "test the use case/adapter/route", or any testing task in a Python backend. Use when testing any Python or FastAPI code.
    3 repo stars
  31. ▌
    Async React Patterns · soludevtech bundle
    React/TypeScript asynchronous patterns: Suspense, the use() hook, server components, streaming SSR, TanStack Query data fetching, concurrent features (useTransition, useDeferredValue), async error boundaries, and testing async components. Trigger on: Suspense, useTransition, useDeferredValue, TanStack Query, server components, streaming SSR, async error boundaries, "async data", "loading state", "streaming", or any async/concurrency task in a React app. Use whenever the task touches async data flow or concurrency in React.
    3 repo stars
  32. ▌
    Documentation Writer · soludevtech
    Use to write the README.md documentation for a projet, Invoke it every time you change the code on a project.
    3 repo stars
  33. ▌
    Linkedin Post Writer · soludevtech bundle
    Draft a new LinkedIn post from scratch using a 2026 hook formula (anaphora, R.I.P., year-pivot, time-anchor, self-proving, paid-vs-free, curiosity-gap, odd-money, contrarian). Runs the humanizer pass and schedules via Publora on approval. Use when the user asks to write a post, needs a hook, or wants a proven format. Not for reviewing existing drafts (use linkedin-humanizer --mode audit).
    3 repo stars
  34. ▌
    Async Nestjs Patterns · soludevtech bundle
    NestJS/TypeScript asynchronous patterns: async/await vs RxJS Observables interop, async interceptors/pipes/guards, event-driven design with @nestjs/event-emitter, background jobs with Bull queues, microservices transport (TCP/Redis/NATS), WebSocket gateways, and lifecycle hooks. Use when implementing async flows, event-driven workflows, queues, microservices, or real-time WebSocket handlers in a NestJS backend.
    3 repo stars
  35. ▌
    Async Python Patterns · soludevtech bundle
    Comprehensive guidance for implementing asynchronous Python applications using asyncio, concurrent programming patterns, and async/await for building high-performance, non-blocking systems. Trigger on: asyncio, async/await, FastAPI background tasks, "async flow", "concurrency", "non-blocking", event loop, "race condition", gather/TaskGroup, or any async/concurrency task in a Python backend. Use whenever the task touches async or concurrency in Python.
    3 repo stars
  36. ▌
    Promptfoo Redteam LLM · soludevtech bundle
    Red team an LLM application end-to-end with promptfoo — the foundational workflow. Use when setting up the first red team for an LLM app, configuring prompts/targets/plugins/strategies, generating adversarial test cases, and reviewing the vulnerability report. Covers OWASP LLM01/08/09, PII, harmful content, hallucination, hijacking.
    3 repo stars
  37. ▌
    Promptfoo Redteam RAG · soludevtech bundle
    Red team Retrieval-Augmented Generation (RAG) applications with promptfoo. Use when testing RAG systems for prompt injection via retrieved context, context injection from compromised knowledge base documents, data poisoning, source attribution fabrication, data/PII exfiltration, context window overflow, and component-level (retrieval vs generation) isolation.
    3 repo stars
  38. ▌
    Feature Implementation · soludevtech
    Skill-driven development workflow for implementation tasks. Use this skill when the user asks to implement a feature, fix a bug, or make significant code changes. Aggressively loads the matching skill via the `skill` tool at EVERY step — TDD, clean architecture, code review, simplification, linting, security, QA, and documentation. Skills version (no agent delegation).
    3 repo stars
  39. ▌
    Hexagonal React Patterns · soludevtech bundle
    React/TypeScript frontend patterns: hexagonal architecture, SOLID, KISS, domain pure-TS, application/hooks, infrastructure/adapters, ports, CVA variants, Zod-in-domain. Trigger on: React, Next.js, TypeScript frontend, package.json with react, *.tsx, vite.config.ts, or any request to implement/refactor/fix/scaffold a component, hook, page, or adapter. Use when working on ANY React/TypeScript frontend, not only explicitly hexagonal ones; defaults the structure to hexagonal.
    3 repo stars
  40. ▌
    Linkedin Content Planner · soludevtech bundle
    Generate a 7-day LinkedIn content plan from a theme, audience, and pillars. Produces per-day post pillar, format, hook type, CTA, posting time, daily comment targets, and a weekly inbound-readiness check. Use when the user wants to plan a week or month of content, not draft a single post.
    3 repo stars
  41. ▌
    Promptfoo Framework Gdpr · soludevtech bundle
    GDPR Data Protection Testing
    3 repo stars
  42. ▌
    Promptfoo Redteam Agents · soludevtech bundle
    Red team LLM agents, MCP servers, and multi-turn chatbots with promptfoo. Use when testing agents for privilege escalation (RBAC/BOLA/BFLA), memory poisoning, multi-stage attack chains, tool/API manipulation, objective hijacking, MCP tool poisoning, tool shadowing, cross-server attacks, and multi-turn conversation manipulation. Covers layered testing (black-box/component/trace-based), OpenTelemetry tracing, trajectory assertions, and stateful strategies.
    3 repo stars
  43. ▌
    Hexagonal Nestjs Patterns · soludevtech bundle
    NestJS/TypeScript backend with hexagonal architecture, SOLID, and KISS. Project structure, ports as abstract classes (inbound/outbound split), Zod entity/DTO validation, injection tokens, exception filters, Swagger via zod-openapi. Use when implementing or refactoring a NestJS backend following hexagonal architecture.
    3 repo stars
  44. ▌
    Hexagonal Python Patterns · soludevtech bundle
    FastAPI / Python backend patterns: hexagonal architecture, SOLID, KISS, ports (ABC inbound/outbound), Pydantic entities, use cases, centralized errors/logging. Trigger on: FastAPI, Python backend, pyproject.toml, uv.lock, or any request to implement/refactor/fix/scaffold a Python endpoint, use case, or adapter. Use when working on ANY FastAPI or Python backend, not only explicitly hexagonal ones; defaults the structure to hexagonal.
    3 repo stars
  45. ▌
    Loop Implementation Review · soludevtech
    Skill-driven implementation loop wrapping feature-implementation. Aggressively loads the matching skill via the `skill` tool at EVERY step. Adds mandatory NEW e2e QA tests in `soludev-compose-apps/<app_name>/e2e` (real path on disk, NO leading `@`), a zero-critical-issues code review gate, one draft PR per modified repo, and a reviewer loop until 0 critical issues and score >= 8/10. Use when the user asks to implement a feature/evolution/bugfix and loop until QA, code review, and PR reviewer sign-off are all green. Skills version (no agent delegation).
    3 repo stars
  46. ▌
    Promptfoo Strategies Static · soludevtech bundle
    Static (deterministic, single-turn) red team strategies for promptfoo — base64, hex, ROT13, leetspeak, homoglyph, image/audio/video encoding, jailbreak-templates, camelCase, emoji smuggling, morse code, pig latin. Use when testing filter bypass via encoding, character substitution, known jailbreak templates, or multimodal encoding without an LLM-as-attacker.
    3 repo stars
  47. ▌
    Feature Implementation Light · soludevtech
    Lightweight skill-driven development workflow for implementation tasks. Use this skill when the user asks to implement a feature, fix a bug, or make code changes but wants a fast loop. Only 4 steps — TDD (Red), implementation (Green + full test suite), code review (0 critical, score ≥ 8/10), and code simplification. Skills version (no agent delegation). No linter, Sonar, Trivy, QA, docs, or PR steps.
    3 repo stars
  48. ▌
    Promptfoo Redteam Guardrails · soludevtech bundle
    Test and validate LLM guardrails (Azure Content Filter, AWS Bedrock Guardrails, NVIDIA NeMo, OpenAI moderation, Google Model Armor) with promptfoo. Use when benchmarking content filters, measuring true/false positives, testing evasion resistance, configuring Model Armor with Vertex AI, or comparing guardrail vendors.
    3 repo stars
  49. ▌
    Promptfoo Redteam Multimodal · soludevtech bundle
    Red team vision, audio, and video LLMs with promptfoo. Use when testing multimodal models against image-embedded harmful text, audio-delivered harmful content, video-embedded attacks, static image + variable text, real unsafe images (UnsafeBench), or VLGuard dataset. Covers the critical injectVar setting, image/audio/video strategies, and dataset plugins.
    3 repo stars
  50. ▌
    Promptfoo Strategies Dynamic · soludevtech bundle
    Dynamic (iterative, single-turn) red team strategies for promptfoo — jailbreak, jailbreak:composite, jailbreak:meta, jailbreak:tree, best-of-n, GCG, citation, likert, math-prompt, authoritative-markup-injection. Use when an LLM-as-attacker iteratively refines prompts to bypass security controls, building attack taxonomies, chaining research techniques, or using academic framing.
    3 repo stars
  51. ▌
    Feature Implementation Agents · soludevtech
    Agent-driven development workflow for implementation tasks. Use this skill when the user asks to implement a feature, fix a bug, or make significant code changes. Aggressively delegates EVERY step to a dedicated agent via the `task` tool — the orchestrator injects a SKILL MANDATE (explicit skill-loading instruction with exact skill names) at the top of every task prompt, because subagents do NOT auto-load skills. Agents version (no direct skill loading by the orchestrator).
    3 repo stars
  52. ▌
    Promptfoo Framework Eu AI Act · soludevtech bundle
    EU AI Act
    3 repo stars
  53. ▌
    Promptfoo Framework Iso 42001 · soludevtech bundle
    ISO 42001
    3 repo stars
  54. ▌
    Promptfoo Framework Owasp API · soludevtech bundle
    Test LLM applications with API access against the OWASP API Security Top 10 (2023) with promptfoo. Use when testing BOLA (API1), broken authentication (API2), excessive data exposure (API3), resource consumption (API4), BFLA (API5), sensitive business flows (API6), SSRF (API7), security misconfiguration (API8), inventory management (API9), or unsafe API consumption (API10).
    3 repo stars
  55. ▌
    Promptfoo Framework Owasp LLM · soludevtech bundle
    Test LLM applications against the OWASP LLM Top 10 (2025) with promptfoo. Use when testing for prompt injection (LLM01), sensitive information disclosure (LLM02), supply chain vulnerabilities (LLM03), data/model poisoning (LLM04), improper output handling (LLM05), excessive agency (LLM06), system prompt leakage (LLM07), vector/embedding weaknesses (LLM08), misinformation (LLM09), or unbounded consumption (LLM10).
    3 repo stars
  56. ▌
    Promptfoo Redteam Multi Input · soludevtech bundle
    Red team multi-input LLM applications with promptfoo — apps that combine user identity, session context, form fields, and messages into one request. Use when testing authorization bypass + prompt injection combos across multiple fields, role confusion attacks, BOLA/BFLA/RBAC with user_id + action inputs, typed DOCX/PDF/image upload workflows, and role-based context testing.
    3 repo stars
  57. ▌
    Promptfoo Redteam Supply Chain · soludevtech bundle
    Secure the LLM supply chain and detect model drift with promptfoo. Use when establishing behavioral baselines, setting up CI/CD pre-deployment security gates, detecting model drift via ASR thresholds, running vendor acceptance tests, comparing fine-tuned vs base models, integrating static (ModelAudit) + dynamic (red team) gates, or alerting on security regressions over time. Covers OWASP LLM03 supply chain vulnerabilities.
    3 repo stars
  58. ▌
    Promptfoo Framework Mitre Atlas · soludevtech bundle
    MITRE ATLAS
    3 repo stars
  59. ▌
    Promptfoo Framework Nist AI Rmf · soludevtech bundle
    Test AI systems against the NIST AI Risk Management Framework (AI RMF) with promptfoo. Use when measuring AI risks across the four core functions (Govern, Map, Measure, Manage), testing specific MEASURE categories (safety, security, privacy, fairness, misuse), or demonstrating NIST AI RMF compliance for federal AI guidelines.
    3 repo stars
  60. ▌
    Promptfoo Redteam Configuration · soludevtech bundle
    Skill: promptfoo-redteam-configuration
    3 repo stars
  61. ▌
    Promptfoo Strategies Multi Turn · soludevtech bundle
    Multi-turn (conversational) red team strategies for promptfoo — crescendo, goat, goblin, hydra, mischievous-user. Use when testing stateful chatbots or agents that maintain conversation history, gradually escalating harm over multiple turns, branching across conversation paths with persistent memory, or simulating persistent creative users.
    3 repo stars
  62. ▌
    Loop Implementation Review Agents · soludevtech
    Agent-driven implementation loop wrapping feature-implementation-agents. Aggressively delegates role steps to dedicated agents via the `task` tool (the orchestrator injects a SKILL MANDATE with exact skill names at the top of every task prompt — subagents do NOT auto-load skills) AND aggressively loads pure-skill steps (code-reviewer, code-simplifier, linter, sonarfix, trivyfix, documentation-writer, githubpr) via the `skill` tool directly. Adds mandatory NEW e2e QA tests in `soludev-compose-apps/<app_name>/e2e` (real path, NO leading `@`), a zero-critical-issues code review gate, one draft PR per modified repo, and a reviewer loop until 0 critical issues and score >= 8/10. Use when the user asks to implement a feature/evolution/bugfix and loop until QA, code review, and PR reviewer sign-off are all green — using agents as the execution layer for roles and skills as the execution layer for tooling steps.
    3 repo stars
  63. ▌
    Promptfoo Framework Dod AI Ethics · soludevtech bundle
    Test AI systems against the DoD AI Ethical Principles (Responsible, Equitable, Traceable, Reliable, Governable) with promptfoo. Use when measuring ethical and security risk for DoD-aligned AI systems, testing human accountability, bias minimization, auditability, safety, and governability.
    3 repo stars
  64. ▌
    Promptfoo Framework Owasp Agentic · soludevtech bundle
    Test AI agent applications against the OWASP Top 10 for Agentic Applications (ASI01-ASI10) with promptfoo. Use when testing agent goal hijack, tool misuse, identity/privilege abuse, agentic supply chain, unexpected code execution, memory/context poisoning, insecure inter-agent communication, cascading failures, human-agent trust exploitation, or rogue agents.
    3 repo stars
  65. ▌
    Promptfoo Redteam Foundation Models · soludevtech bundle
    Assess foundation and fine-tuned LLM security with promptfoo — dynamic red team scans against live models plus static ModelAudit scanning of model files. Use when baseline-testing a base/foundation model, comparing multiple models side-by-side, scanning model files for trojans/malicious pickle/embedded executables, running HarmBench standardized benchmarks, or contributing results to promptfoo.dev/models.
    3 repo stars
  66. ▌
    Promptfoo Strategies Custom Regression · soludevtech bundle
    Custom, composition, and regression red team strategies for promptfoo — custom (text-based multi-turn), custom strategy scripts (JavaScript), layer (compose multiple strategies sequentially), retry (regression testing from historical failures). Use when creating user-defined attack strategies, chaining strategies together (jailbreak → encoding, agentic → multimodal), building regression test suites from past failures, or writing custom JavaScript strategy transforms.
    3 repo stars
  67. ▌
    Loop Implementation Review Agents Light · soludevtech
    Lightweight agent-driven implementation loop wrapping feature-implementation-light. Aggressively delegates role steps to dedicated agents via the `task` tool (the orchestrator injects a SKILL MANDATE with exact skill names at the top of every task prompt — subagents do NOT auto-load skills) AND loads the pure-skill step (code-simplifier) via the `skill` tool directly. Only 4 steps — TDD, implementation, code review (0 critical + score >= 8/10 with reviewer loop), and simplification. No QA, lint, Sonar, Trivy, docs, or PR steps. Use when the user asks to implement a feature/evolution/bugfix with a fast loop until code review sign-off — using agents as the execution layer for roles and skills as the execution layer for tooling steps.
    3 repo stars
  68. ▌
    Promptfoo Strategies Indirect Injection · soludevtech bundle
    Indirect prompt injection strategy for promptfoo — indirect-web-pwn. Use when testing AI agents with web browsing capabilities for manipulation via malicious instructions embedded in web pages they fetch. Tests data exfiltration (with data-exfil plugin, deterministic server-side tracking) and behavior manipulation (with any other plugin, LLM-based grading). Supports layering with jailbreak strategies for more effective attacks.
    3 repo stars