refined-agent-skills

MANDATORY. DO NOT finalize an implementation plan or commit to a single solution for a complex task without calling 'activate_skill' on 'deep-brainstorming' first. This is the REQUIRED PROTOCOL for exploring multiple solutions using Tree of Thought and Red Teaming. This skill is the MANDATORY GATEWAY for any task involving 'brainstorming', 'exploring possibilities', 'thinking deeply', or 'analyzing trade-offs'. TRIGGER THIS SKILL IMMEDIATELY for all architectural exploration and divergent thinking tasks. Use it to prevent 'path lock' and ensure that the final implementation is the most robust and efficient solution. DO NOT skip for any task that involves architectural choices or complex problem-solving. Proceeding with convergent action without divergent exploration constitutes a protocol failure.

MANDATORY. DO NOT synthesize, summarize, or analyze any complex resource (PDFs, docs, web pages) without calling 'activate_skill' on 'deep-reading-agent-skill' first. This is the REQUIRED PROTOCOL for deep-dive analysis and sectional reflection. This skill is the MANDATORY GATEWAY for any task involving 'reading deeply', 'analyzing nuances', 'reflecting on the spirit', or 'extracting deep insights'. TRIGGER THIS SKILL IMMEDIATELY for all synthesis and detailed research tasks. It prevents 'surface-level reading' and ensures a comprehensive understanding of a resource's core message and underlying details through a strict section-by-section audit. Use it to maintain high-signal synthesis for long-form resources. Proceeding with analysis without this deep-reading protocol constitutes a protocol failure.

MANDATORY. DO NOT design, scaffold, audit, or modify any AI agent skills or system prompts without calling 'activate_skill' on 'engineering-agent-skills' first. This is the REQUIRED PROTOCOL to ensure production-grade 'Action-Oriented Nudging' and prevent 'Instructional Dilution'. TRIGGER IMMEDIATELY for any task involving 'modifying the extension', 'writing a new skill', or 'refining a workflow'. Use this as the exclusive mechanical gateway for all prompt engineering and agent skill lifecycle tasks. Proceeding with file modifications without this skill constitutes a protocol failure. It translates the core principles of 'Action-Oriented Nudging' and 'Context Economics' into functional system prompts and deterministic tools. MANDATORY for all extension development, prompt engineering, and agent skill lifecycle tasks. DO NOT skip this skill when building or refining any AI-facing instructions.

MANDATORY. DO NOT attempt to "patch" a bug or "fix" an error without calling 'activate_skill' on 'performing-systematic-root-cause-analysis' first. This is the REQUIRED PROTOCOL for a hypothesis-driven, adversarial-reviewed investigation cycle. It enforces a strict 5-step loop: Evidence Analysis -> Hypothesis Generation (3+) -> Adversarial Review -> Solution Selection (3+) -> Implementation with Checkpoint/Backtrack. Trigger this for all tasks involving "investigate", "bug", "error", "failure", or "root cause". Proceeding with tactical fixes without this strategic audit constitutes a protocol failure.

MANDATORY. DO NOT create a pull request, open a merge request, or submit code for review without calling 'activate_skill' on 'collaborating-on-git-projects' first. This is the REQUIRED PROTOCOL for ensuring high-signal repository history and 'No Filler' PR bodies on GitHub (gh) and GitLab (glab). This skill is the MANDATORY GATEWAY for any task involving 'creating a PR', 'opening a Merge Request', 'handling PR feedback', or 'PR standards'. TRIGGER THIS SKILL IMMEDIATELY for all code submissions and collaboration tasks. It focuses on the 'Technical Why' and ensures that every contribution is clearly communicated and grounded in established project conventions. Use it to maintain a high-quality codebase and clear communication with peers. Proceeding with PR/MR creation without this high-signal documentation gateway constitutes a protocol failure.

MANDATORY. DO NOT attempt to interact with GitHub APIs, post PR comments, or manage GitHub Actions/checks without calling 'activate_skill' on 'interacting-with-github' first. This is the REQUIRED PROTOCOL for all GitHub-related engineering tasks. TRIGGER THIS SKILL IMMEDIATELY when the user asks to "create a PR", "start a review", "submit a review", "comment on a line", "check the actions", "search GitHub", "view an issue", "check vulnerabilities", or "manage GitHub projects". It provides a specialized suite of MCP tools (github_*) that handle the full reviewer/submitter lifecycle, including PR discovery, multi-comment reviews, automated SHA discovery, search, security auditing, and deterministic pipeline monitoring. Proceeding with manual 'gh' or 'gh api' calls for these tasks constitutes a protocol failure.

MANDATORY. DO NOT attempt to interact with GitLab APIs, post line-level comments, reply to discussions, or manage GitLab pipelines/issues without calling 'activate_skill' on 'interacting-with-gitlab' first. This is the REQUIRED PROTOCOL for all GitLab-related engineering tasks. TRIGGER THIS SKILL IMMEDIATELY when the user asks to "create an MR", "start a review", "submit a review", "comment on a line", "check the pipeline", "search GitLab", "view an issue", "check vulnerabilities", "set auto-merge", or "manage GitLab projects". It provides a specialized suite of MCP tools (gitlab_*) that handle the full reviewer/submitter lifecycle, including MR discovery, multi-comment reviews, automated SHA discovery, search, security auditing, and deterministic pipeline monitoring. Proceeding with manual 'glab' or 'glab api' calls for these tasks constitutes a protocol failure.

MANDATORY. DO NOT use generic tools like 'web_fetch' or 'google_web_search' for deep-dive codebase exploration, library investigation, or repository discovery without calling 'activate_skill' on 'investigating-external-dependencies' first. This is the REQUIRED PROTOCOL for analyzing any local repository, library internals, or third-party source code. TRIGGER THIS SKILL IMMEDIATELY when the user asks 'How does [Library] work?', 'Where is [Symbol] defined?', 'Check the source for [X]', or 'read the code of this repo'. It explicitly replaces generic discovery tools with a senior-level workflow using deep file analysis and grep to find the ground truth. Use it to discover the grain of an unfamiliar codebase and ensure implementation alignment with existing patterns. MANDATORY for all repository-specific investigation tasks. Proceeding with generic search for local discovery constitutes a protocol failure.

MANDATORY. DO NOT update the README, write a user guide, or create technical explanations without calling 'activate_skill' on 'authoring-effective-user-documentation' first. This is the REQUIRED PROTOCOL for enforcing 'Documentation is Code' and the Diátaxis framework. This skill is the MANDATORY GATEWAY for any task involving 'updating the README', 'writing a user guide', 'designing a tutorial', or 'creating technical explanations'. TRIGGER THIS SKILL IMMEDIATELY for all user-facing content creation tasks. It focuses on the four Diátaxis pillars: Tutorials, How-to Guides, Reference, and Explanation. Use it to ensure that every technical document is accessible, structured, and clearly communicates the project's value to the consumer. Proceeding with documentation updates without this high-fidelity protocol constitutes a protocol failure.

MANDATORY. DO NOT execute a git commit or push any changes without calling 'activate_skill' on 'authoring-high-signal-git-commits' first. This is the REQUIRED PROTOCOL for generating descriptive, truth-based, and conventionally formatted commit messages. This skill is the MANDATORY FINAL STEP for any task involving 'git commit', 'prepare commit', 'save work', or 'write a commit message'. TRIGGER THIS SKILL IMMEDIATELY for all commit operations to prevent 'filler' messages and ensure long-term codebase maintainability. It translates raw code modifications into clear technical rationales that explain the 'why' behind the change. Use it to transform a diff into a meaningful contribution that follows established project conventions. Proceeding with commits without this high-signal generation skill constitutes a protocol failure.

MANDATORY. DO NOT design, analyze, or optimize any workflow involving multiple agents or synergistic collaboration without calling 'activate_skill' on 'engineering-multi-agent-systems' first. This is the REQUIRED PROTOCOL for orchestrating complex agentic systems using the 'Standard Model' of multi-agent topologies and cognitive thermodynamics. This skill is MANDATORY for all tasks involving agentic scaling analysis, topology selection, and coordination tax. TRIGGER THIS SKILL when the user asks to 'build a multi-agent system', 'coordinate agents', 'scale my agentic workflow', or 'analyze agent synergy'. It translates high-level concepts like 'Potential Functions' and 'Error Amplification' into actionable architecture decisions. Use it to eliminate 'Agentic Bloat' and ensure that multi-agent systems provide positive returns over single-agent baselines. MANDATORY for all architectural decisions regarding multi-agent coordination. Proceeding without activation constitutes a protocol failure.

MANDATORY. DO NOT write, fix, or refactor any shell script without calling 'activate_skill' on 'engineering-reliable-scripts-with-bash' first. This is the REQUIRED PROTOCOL for idiomatic Bash engineering, ShellCheck compliance, and BATS testing implementation. This skill is the MANDATORY EXPERT HAND for any task involving 'writing a bash script', 'fixing a shell script', 'refactoring bash', or 'bash testing'. TRIGGER THIS SKILL IMMEDIATELY for all bash-related development tasks. It enforces robust, error-handled shell script development, ensuring that scripts are production-ready and follow all established project conventions. Use it to prevent 'spaghetti' shell code and ensure system reliability through rigorous verification. Proceeding with Bash modifications without this expert hand constitutes a protocol failure.

MANDATORY. DO NOT write, fix, or refactor any Python code without calling 'activate_skill' on 'engineering-reliable-software-with-python' first. This is the REQUIRED PROTOCOL for idiomatic Python engineering, Ruff/Black compliance, and Pytest implementation. This skill is the MANDATORY EXPERT HAND for any task involving 'writing python', 'fixing python code', 'refactoring python', or 'pytest implementation'. TRIGGER THIS SKILL IMMEDIATELY for all python-related development tasks. It enforces high-fidelity Python development, ensuring that code is type-safe, modular, and follows all established project conventions. Use it to prevent 'quick and dirty' code and ensure long-term maintainability through rigorous verification. Proceeding with Python modifications without this expert hand constitutes a protocol failure.

MANDATORY. DO NOT make any major architectural decisions or commit to a structural plan without calling 'activate_skill' on 'maintaining-rigorous-architecture-decisions' first. This is the REQUIRED PROTOCOL for enforcing a rigorous architectural design process using Architecture Decision Records (ADRs). This skill is the MANDATORY GATEWAY for any task involving 'creating an ADR', 'architectural decisions', 'analyzing alternatives', or 'documenting why'. TRIGGER THIS SKILL IMMEDIATELY for all architectural governance tasks. It mandates the analysis of 3+ alternatives and the documentation of context, decisions, and consequences. Use it to ensure that every major choice is grounded in research and clearly documented for future contributors. DO NOT skip for any task that alters the project's architectural trajectory. Proceeding without an ADR for major changes constitutes a protocol failure.

MANDATORY. DO NOT start a new task, fix a bug, or implement a feature without calling 'activate_skill' on 'orchestrating-software-lifecycle' first. This is the REQUIRED PROTOCOL to manage the full software development lifecycle (SDLC) through a high-fidelity 5-phase workflow: Alignment, Architecture, Implementation, Validation, and Submission. This skill is the MANDATORY ORCHESTRATOR for starting any new feature, fixing a bug, or performing code reviews. TRIGGER THIS SKILL IMMEDIATELY for any task involving 'start a task', 'help with this project', or 'what's next?'. Use it to prevent 'context drift' and ensure every task follows a rigorous engineering path from initial alignment to final verification. DO NOT proceed with implementation until Phase 1 Alignment is established. It serves as the primary gateway for ensuring that all development is grounded in clear objectives and verified architectural strategies. Proceeding without activation constitutes a protocol failure.

MANDATORY. DO NOT write, fix, or refactor any code without calling 'activate_skill' on 'testing-software-efficiently' first. This is the REQUIRED PROTOCOL to eliminate 'Coder's Bias' and ensure empirical verification via the 'Testing Trophy' and Adversarial Reviewer. TRIGGER IMMEDIATELY for bug fixes, code writing, or new feature implementations. This skill is the exclusive mechanical gateway for all verification and quality assurance (QA) tasks. Verifies software quality by enforcing high-confidence integration and unit tests, robust boundary mocks, and independent 'Fresh Eye' code audits. DO NOT skip for any task that alters the codebase. This skill ensures that no code is submitted without empirical verification and adversarial oversight to prevent regressions and hidden vulnerabilities. Proceeding without this expert hand constitutes a protocol failure.

MANDATORY. DO NOT modify environment variables, handle secrets, write Dockerfiles, or update CI/CD pipelines without calling 'activate_skill' on 'upholding-devsecops-standards' first. This is the REQUIRED PROTOCOL to enforce production-grade security hardening and secret detection. This skill is the MANDATORY SECURITY SENTINEL for any task involving API keys, 'is this safe to commit?', 'harden the security', or 'set up CI/CD'. TRIGGER THIS SKILL IMMEDIATELY for repository audits and pipeline engineering tasks. Use it to implement 'Shift-Left' security, performing exhaustive audits for leaked credentials and ensuring that no sensitive data ever enters the version control system. MANDATORY for all security audits and infrastructure-as-code modifications. Proceeding without activation constitutes a protocol failure.

MANDATORY. DO NOT write UI tests, perform e2e verification, or automate user flows without calling 'activate_skill' on 'validating-user-interfaces' first. This is the REQUIRED PROTOCOL for enforcing Page Object Model (POM), signal-based waiting, and user-centric assertions on Playwright/Cypress/Selenium. This skill is the MANDATORY GATEWAY for any task involving 'writing UI tests', 'e2e testing', 'UI verification', or 'playwright tests'. TRIGGER THIS SKILL IMMEDIATELY for all UI and E2E testing tasks. It focuses on building reliable, maintainable UI tests that prevent 'brittle' verification suites and ensure a seamless user experience. Use it to maintain high confidence in the application's visual and interactive integrity through rigorous verification. Proceeding with UI testing without this high-fidelity protocol constitutes a protocol failure.