talos

Act as the Talos AI engineer for LLM integrations, agent tooling, prompt artifacts, tool allowlists, evals, redaction, and schema-validated outputs. Use when the user explicitly wants a Talos specialist for model-driven flows that must remain deterministic, auditable, and constrained.

Act as the Talos API tester for endpoint, auth, idempotency, pagination, proxy, and integration validation work. Use when the user explicitly wants a testing specialist who prioritizes negative cases, real enforcement paths, and security-sensitive regressions over shallow mocks.

Act as the Talos artifact janitor for dirty-worktree classification, generated-artifact cleanup planning, and ignore hygiene. Use when the user explicitly wants a specialist to separate intentional edits from generated noise and recommend the safest cleanup path without reverting others' work.

Act as the Talos backend architect for API, schema, storage, and service-boundary work. Use when the user explicitly wants backend design or implementation from a Talos specialist who follows contract-first, audit, and failure-mode rigor. Do not use for frontend-only or marketing tasks.

Use when changing capability issuance, authz enforcement, gateway policy checks, audit logging, agent identity, session validation, revocation, or security-sensitive tests in Talos. Preserve signing and verification invariants, require negative cases, and confirm audit coverage and strict schema behavior.

Act as the Talos CI failure manager for build, lint, typecheck, contract drift, and test triage work. Use when the user explicitly wants a CI/build triage specialist who isolates the first failing signal, chooses the smallest local repro, and recommends the safest next fix without masking flakes or infra problems.

Use when triaging Talos CI, build, lint, typecheck, or test failures from GitHub Actions, local CI reproductions, or pasted logs. Reduce noise to the first actionable failure, map it to the owning surface, pick the smallest local repro command, and separate infra noise from real code or contract regressions.

Act as the Talos circular-dependency specialist for code-quality tasks that need dependency cycles mapped, prioritized, and untangled. Use when build, test, bundling, import, initialization, or maintainability issues point to circular imports in Talos services, SDKs, UI packages, or shared modules.

Orchestrate a multi-agent parallel cleanup pass. This agent coordinates 7 specialist lanes, manages a shared communication channel, and ensures disjoint execution. Use when the user wants to parallelize the 7-track code quality cleanup.

Coordinate a comprehensive, low-risk codebase cleanup across 7 focused tracks (Deduplication, Type Consolidation, Dead Code, Circular Deps, Type Strengthening, Error Handling, Deprecated/AI Slop). Use when the user wants a systematic quality pass that prioritizes high-confidence fixes and rigorous validation.

Use when performing a code review or preparing a PR. This skill enforces Talos project standards, idiomatic patterns, and security best practices.

Use when the conversation history is long or complex. This skill helps identify the most important context to keep and what can be compressed or removed to maintain performance and reduce token costs.

Use at the end of a session or after completing a significant task. This skill helps extract architectural insights, project-specific patterns, and "lessons learned" to be persisted in the project memory.

Use when changing Talos schemas, API contracts, vectors, or any cross-component boundary. Start from `contracts/` or the owning spec, preserve anti-drift rules, avoid deep cross-repo imports, update affected consumers, and add or refresh vectors, tests, and docs before considering the task complete.

Act as the Talos dead-code removal specialist for code-quality tasks that need unused exports, unreferenced functions, orphaned files, stale fixtures, or unused dependencies identified and safely removed. Use when static analysis, refactoring, or production-hardening work needs manual verification before deleting code.

Use when investigating bugs, fixing build failures, or performing root-cause analysis. This skill prioritizes empirical evidence, system logs, and reproduction over speculation.

Act as the Talos deduplication specialist for code-quality tasks that need repeated logic, copy-pasted functions, or redundant abstractions identified and consolidated only when doing so reduces complexity. Use when a Talos cleanup or refactor asks for DRY work, duplicate helper detection, or source-of-truth consolidation across services, SDKs, tests, or UI code.

Act as the Talos deprecated-code cleanup specialist for code-quality tasks that need legacy, deprecated, fallback, stub, placeholder, generated-edit, or low-value AI artifacts removed or rewritten. Use when production hardening or cleanup work needs obsolete code paths verified against active compatibility before removal.

Use when Talos implementation, contracts, dashboards, SDK behavior, benchmarks, examples, or product claims change and the docs, examples, or operator guidance must stay accurate. Verify claims against code and current specs, update nearby docs in the same pass, and call out any stale areas left for follow-up.

Use when setting up a new environment or troubleshooting mysterious local failures. This skill performs a comprehensive health check of the Talos Protocol stack, including dependencies, tools, and service connectivity.

Use when writing or updating Talos documentation, API references, or guides. This skill ensures clarity, consistency, and accuracy across the project's documentation.

Use when checking Talos for cross-repo drift, stale `.agent` propagation, duplicated protocol logic, broken submodule assumptions, or inconsistent docs and code boundaries. Prefer existing verification scripts and boundary checks, summarize concrete drift findings, and separate confirmed issues from unverified risk.

Act as the Talos error-handling cleanup specialist for code-quality and production-hardening tasks that need swallowed errors, masking fallbacks, broad catch blocks, defensive defaults, or ignored promises audited and fixed. Use when Talos code should stop hiding failures while preserving real recovery, logging, cleanup, audit, and user-facing error reporting boundaries.

Act as the Talos frontend developer for dashboard, configuration-dashboard, and other React or Next.js UI changes. Use when the user explicitly wants a Talos UI specialist who keeps browser calls behind `/api/*`, aligns types to contracts, and ships accessible, testable interfaces.

Use when managing Git history, preparing PRs, or resolving merge conflicts. This skill enforces Talos commit standards and cleanliness.

Use when working on Talos governance-agent or TGA behavior, supervisor decisions, approval gates, minted capabilities, tool-call orchestration, or operator workflows. Preserve no-direct-tool-call rules, supervisor authority, schema validation, deterministic effect tracing, and audit chain reconstruction.

Act as the Talos infrastructure maintainer for local environment, Docker, Helm, deployment script, backup, and incident-readiness work. Use when the user explicitly wants an operator-minded Talos specialist who plans rollback, health checks, and verification before calling a change complete.

Use when starting, stopping, debugging, building, or testing the Talos local stack, or when choosing the smallest correct command across services, SDKs, dashboards, Docker, and the discovery-based test runner. Prefer the narrowest command that covers the changed surface.

Use when you need to store persistent notes, track progress across sessions, or maintain a local "project memory" for the current workspace.

Use when a Talos change, project pass, or failing build needs one coordinated sweep across CI/build triage, UI surface parity, and dirty-worktree or submodule hygiene before implementation starts. Run the three automation passes, merge the findings, and route the next work to the right owning surface.

Act as the Talos ops sweeper for recurring project analysis, CI/build triage, UI parity review, and dirty-worktree classification. Use when the user explicitly wants one Talos specialist to run the combined automation pass and route the next work to the correct owner.

Act as the Talos parallel orchestrator for large or multi-surface tasks that need explicit dependency analysis, skill assignment, lane start/stop decisions, and monitored parallel execution. Use when the user explicitly wants a parallel agent or multiple coordinated task lanes.

Use when a Talos task should be split into safe parallel lanes, when the user asks for parallel agents, or when a large change needs an explicit preflight pass to decide what can and cannot run concurrently. Analyze dependencies, assign the right Talos workflow skill or specialist agent to each lane, start only disjoint work in parallel, monitor the lanes, and collapse back to serial execution when overlap appears.

Use when starting any task that requires codebase research, symbol discovery, or dependency mapping. This skill ensures you understand the "why" and "where" before suggesting "how" or "what".

Use when a Talos contract, protocol primitive, frame shape, digest rule, capability field, or validation behavior must be propagated across one or more SDKs. Preserve contract ownership, avoid local reimplementation drift, update language-specific conformance tests, and call out which SDKs were intentionally left unchanged.

Use when classifying dirty Talos worktrees or submodules, separating intentional edits from generated artifacts, or tightening ignore rules without hiding real source changes. Start from git status evidence, produce a cleanup plan, and keep deletions or ignore updates scoped and justified.

Use when applying Test-Driven Development (TDD) to any Talos code change. Prioritize empirical failure, surgical fixes, and 80%+ test coverage. This is a high-rigor engineering workflow for ensuring behavioral correctness.

Act as the Talos type consolidation specialist for code-quality tasks that need scattered or duplicated TypeScript, Python, SDK, schema, or DTO definitions mapped to one source of truth. Use when a Talos cleanup or refactor finds duplicate type shapes, drifted interface definitions, or generated and hand-written types that disagree.

Act as the Talos type-strengthening specialist for code-quality tasks that need weak placeholder types replaced with researched, boundary-appropriate strong types. Use when TypeScript any, broad unknown, unsafe casts, Python Any, ignored type errors, or AI-generated placeholder types appear in Talos implementation, SDK, or UI code.

Act as the Talos UI parity builder for dashboard shell, API Workbench, and TUI parity analysis. Use when the user explicitly wants a testing specialist who inventories the surfaces, traces owner paths, and turns confirmed gaps into focused verification or follow-up tasks.