envoydev
- 129 skills
- 0 followers
- 1 repo stars
- 3 days ago last updated
- ▌ Project Implementer · envoydevUse when you have a task plan in hand - from project-solution-design, ideally gated by project-verify-plan - and want to BUILD it in the current chat, task by task: the single-chat form of the implementer seat's execution protocol. Honors each task's contract, builds code + tests per task, gates each task green before the next, resolves every red INLINE (in this chat, dispatching nothing), and hands to the build review (`project-verify-code` inline or the `<stack>-verifier` seat) + the done-gate. Trigger on execute the plan, build the plan, implement the tasks, build task 2, continue the plan. On an agents answer (the run-start ask, or the calling flow's recorded mode) it hands each task to its `<stack>-implementer` seat instead (up to 3 at once, each on its frontmatter model unless you name one); the full multi-agent flow with its own designer and verifier is project-solve-cross-task. Not a plan-less ad-hoc edit - just make that.
- ▌ Project Verify Code · envoydevUse when a build is assembled and you want to review it in THIS chat with no dispatch - the single-chat, no-agents form of the verifier seat, and the inline alternative to /code-review's fan-out. Loads the stack's trap-list skills, reruns build + tests, gates the code against its plan, RUNS the app on failable inputs (a test can pass under WebApplicationFactory while the live endpoint 500s), traces a changed wire contract to its consumers, and returns a ranked punch-list - all inline, dispatching nothing. Trigger on review the build, review this here, check the code without agents, review before done, verify the build. Not the plan audit (project-verify-plan, before code), not the dispatched verifier seat or /code-review's parallel angles - this is the review you run without spawning anything.
- ▌ Project Verify Plan · envoydevUse when you have an implementation plan or design in hand and want to audit it BEFORE writing code - a risk-coverage review that checks the plan names the non-obvious traps its stack will actually hit, matches the requirement's scope, covers the edge and safety cases, and stays minimal. The cheapest place to catch a design error, since a flawed plan built perfectly is still wrong. Pairs with writing-plans (which creates the plan) and precedes project-verify-code (which reviews the built code). Trigger on review this plan, is this design sound, does the plan miss anything, before I build.
- ▌ Database Conventions · envoydev bundledatabase conventions across Postgres, SQL Server/T-SQL, SQLite, and MongoDB - the engine-neutral rules for schema design, migrations, indexes, foreign keys, transactions, connection management, query safety, N+1 prevention, and secret handling, plus the per-engine pitfalls that bite. Load before designing or modifying a schema, writing SQL raw or through an ORM, modeling a document store, or creating a migration, view, procedure, or index. Deeper work routes out: engine tuning to `postgres` / `sqlite`, .NET data access to `dotnet-data-access`, migration mechanics to `dotnet-migrate`, security posture to `database-security`. Do NOT load for app-only in-memory data structures or a project with no persistence layer.
- ▌ Dotnet Project Setup · envoydev bundleSet up a new .NET solution's build spine - the canonical src / tests / .config layout, .slnx solution files, Directory.Build.props shared build properties, global.json SDK pinning + rollForward, central package management via Directory.Packages.props, and pinning a dotnet tool in .config/dotnet-tools.json. Load to set up a new .NET solution, add a NuGet package, add a project, or pin a dotnet tool; trigger files .slnx, Directory.Build.props, Directory.Packages.props, global.json, .config/dotnet-tools.json. Companions, where the project installed them: the .NET router, the .NET analyzer/quality-gate skill, the CI-and-deploy skill; schema and version migrations belong to the EF migration skill. Do NOT load for analyzers / TreatWarningsAsErrors / .editorconfig (the .NET quality-gate skill) or CI workflows / packaging / SourceLink (the CI-and-deploy skill).
- ▌ Project Quality Loop · envoydev bundleAutonomous review-and-fix pipeline driven from a folder of numbered prompt files (the `<docs-path>/loops/` folder of fix-discipline / structure / code-quality / naming / logging / comments). Runs each prompt in numeric order, looping it on a target until its bar is met, then advances - making and logging judgment calls itself, pausing only at the run-start mode ask and each stage-close fresh-session ask. Triggers on 'run the project quality loop' or 'run the loops pipeline'. Code-quality only - architecture restructuring is project-architecture-quality-loop; measuring or raising test coverage is the project-test-coverage-analyzer/-loop pair; a single diff sweep is project-verify-code or /security-review, not this. Do NOT fire for a one-off review pass or when findings should be reported without auto-fixing (a missing or empty loops folder is no blocker - the skill seeds it from a bundled starter set to bootstrap; a folder of custom prompts is the user's, never overwritten).
- ▌ Dotnet Authentication · envoydev bundleASP.NET Core auth conventions covering both halves - authentication (who the caller is) and authorization (what they may do). Pick the scheme by surface: JWT bearer for stateless APIs, cookies for server-rendered apps, OpenID Connect for delegated SSO. Validate every token field, lean on ASP.NET Identity as the user store, and gate access with named policies and authorization handlers rather than scattered role strings. Floors at .NET 8 / C# 12. Load before standing up a sign-in flow, wiring JWT or OIDC, writing an authorization policy, or protecting an endpoint. Companions: csharp, dotnet-minimal-api, dotnet-web-backend, dotnet-cryptography, dotnet-security. Do NOT load for the OWASP hardening sweep or secret placement (dotnet-security) or crypto primitives (dotnet-cryptography).
- ▌ Csharp Design Patterns · envoydev bundleApply GoF design patterns idiomatically in C#/.NET. Use whenever the user asks to implement, choose, explain, or compare a design pattern in C#, refactor toward a pattern, review for pattern misuse, asks 'which pattern fits this problem' - or describes a recurring design problem (object creation sprawl, switch-on-type logic, tight coupling, notification chains, undo/redo, plugin architecture) without naming a pattern. Covers all 23 GoF patterns with modern .NET 8+ idioms; pair with `csharp` for output style. Do NOT use for architectural patterns (repository, unit of work, CQRS, event sourcing, layering) - those route via `dotnet` to `dotnet-architecture`.
- ▌ Dotnet Hosted Services · envoydev bundle.NET hosted-service and worker conventions - the long-running background work the generic host runs. Covers the host shapes (worker binary, in-web-app task, Windows Service), IHostedService versus BackgroundService versus IHostedLifecycleService, the ExecuteAsync exception trap, scoped services from the singleton host, PeriodicTimer over Task.Delay, graceful shutdown, and System.Threading.Channels queues - plus references/ for 24/7 I/O hardening, scheduling/leader election, deployment/signals. Floors at .NET 8 / C# 12. Load when writing a worker service, BackgroundService or IHostedService, a periodic job, a bot/daemon host, or any in-process background task hung off the host. Companions: dotnet-messaging, dotnet-web-backend, csharp. Do NOT load for the broker side of a consumer (dotnet-messaging; the consumer's host process is still this skill), HTTP endpoints, or reactive in-memory streams.
- ▌ Dotnet Mvc Controllers · envoydev bundleASP.NET Core controller-based Web API mechanics - the mainstream, brownfield alternative to minimal APIs. Covers the ApiController attribute, attribute routing, ActionResult of T versus IActionResult versus typed HttpResults, suppressing the automatic 400 filter (ApiBehaviorOptions, SuppressModelStateInvalidFilter) for the house FluentValidation-in-a-filter convention, binding-source inference and explicit From-attributes, IAsyncActionFilter ordering, and thin controllers delegating to services. Floors at .NET 8 / C# 12; 9/10 deltas flagged optional. Load before writing or editing API controllers and action filters. Companions: dotnet-minimal-api, dotnet-web-backend, dotnet-web-error-handling, dotnet-authentication. Do NOT load for minimal APIs, MVC views, Razor Pages, gRPC, SignalR, or non-HTTP code.
- ▌ Dotnet Windows Service · envoydev bundleWindows Service conventions - the Service Control Manager layer over the .NET generic host: AddWindowsService and the dual-mode binary, SCM start/stop budgets, non-zero exit codes so recovery actions actually fire, the System32 working-directory trap, scripted sc.exe install with recovery actions, gMSA / least-privilege service accounts, unquoted-path and ACL hardening, event-log source registration, and the maintained .NET Framework ServiceBase shape + migration path. Load when building, installing, hardening, or migrating a Windows Service - AddWindowsService / UseWindowsService, sc.exe, ServiceBase, installutil, service accounts, SCM errors like 1053. Companions: dotnet-hosted-services (the host model this stacks on), csharp. Do NOT load for the generic worker/host model itself with no SCM target (dotnet-hosted-services), Linux daemons/systemd, or containerized workers.
- ▌ Project Related Context · envoydevThe deliberate related-projects capture: given paths or git URLs to sibling repos, fan out related-project-analyzer agents (one per sibling, parallel) and write BOTH tiers from their entries - the always-on awareness rule .claude/rules/baseline-project-related-context.md (lean name / location / relation / seam per sibling) and <docs-path>/related-context/PROJECT-RELATED-CONTEXT.md (the on-demand orientation doc; its related-context/ folder is the one home for ALL sibling-repo docs - cross-repo plans, change requests, run recipes). Re-run to refresh: entries upserted per passed sibling, unlisted entries kept. Args-driven - it analyzes the locations you name, it never scans for siblings. Triggers on 'capture the related projects' or 'map the sibling repos'. NOT for this repo's own architecture (project-architecture-analyzer) or dynamic cross-repo findings (the memory MCP).
- ▌ Project Solution Design · envoydevUse when you have a feature or change to build in a single chat and want to settle how it fits the existing code before writing any - the single-chat form of the solution-designer seat - orient, judge the fit, decompose into an ordered minimal plan. Trigger on analyse how to integrate this, how does this fit, design this feature, where does this belong, break this into tasks, plan this change. Not for a one-line edit; not the multi-agent flow with its own verifier and fan-out (that is project-solve-cross-task) - in-session it keeps the whole design in your context so you check each step; the run-start ask (or your named mode) decides the designer seat.
- ▌ Project Version Upgrade · envoydev bundleThe deliberate version-upgrade flow for any BREAKING version event - a framework or runtime major, an EOL, a load-bearing package's breaking major: plan in-session (the published breaking-change surface crossed against located usage - applicable changes only), present the staged plan at an approval gate, then drive the execution stage by stage with a green gate after every stage. Auto mode - skipping the approval gate - runs ONLY when the user explicitly asked for it. Routine non-breaking bumps need no skill: just bump. Triggers on 'upgrade to .NET 10', 'ng update to v20', 'this package's new major breaks us', 'plan the framework upgrade'. NOT for a feature that merely needs a newer package (the feature's own flow), or a red CI pipeline (ci-failure-diagnoser, or its single-chat twin project-ci-failure-signatures).
- ▌ Dotnet Source Generators · envoydev bundleRoslyn source-generator conventions, both sides of the line - reaching first for the framework generators that exist (GeneratedRegex, LoggerMessage, the System.Text.Json context) and authoring your own only when none fits; the authoring mechanics live in references/authoring.md. Floors at .NET 8 / C# 12. Load when writing or reviewing a generator, weighing compile-time codegen against reflection, or when the user says source generator, IIncrementalGenerator, ISourceGenerator, GeneratedRegex, LoggerMessage, or ForAttributeWithMetadataName. Companions: csharp for the generator's language style, dotnet-testing for test setup. Do NOT load for ordinary code that ships no generator.
- ▌ Project Diagnose Failure · envoydevUse to investigate a failure of any kind in THIS chat, from whatever evidence you actually have - a pasted Sentry event, one log file, several log sources at once, a red CI run, a stack trace, a screenshot, or nothing but a customer saying checkout is slow. Four gated steps: triage the evidence to a tier, gather (inline or evidence-gatherer seats), prove the root cause, then a user fork - write a report, plan the fix as contracted tasks, or add log points and re-run. Read-only throughout: it never writes the fix. Trigger on investigate this bug, diagnose this failure, what is causing this, look into this incident, triage this report. Not the fix build (project-solve-task takes the tasks from here), not a signature lookup you already have the answer for.
- ▌ Project Solve Cross Task · envoydev bundleThe entry-point router for multi-agent engineering work - scope the task IN-SESSION (the generated awareness rules + a bounded serena pass), then ask session-or-agents up front (an invocation naming the mode is the answer) and route to the smallest safe execution mode: single-chat, one implementer, a single-stack design-build-verify trio, or a producer-first cross-domain run (dispatch recommended in the ask) where the producer designer's interface IS the contract and the integration-reviewer gates the assembled feature. Triggers on how should I build or route this work, plan the agents for this, this spans backend and frontend, or investigate-and-fix a bug across the stack; name the stack in the ask ('frontend only', 'just the API') to pin routing straight to it. This skill scopes and routes - it never designs or writes code, and it runs in the MAIN session only: a dispatched seat never re-fires it, its brief already carries its slice of the flow. NOT for greenfield (project-build-from-scratch) or a deliberat
- ▌ Dotnet Architecture Tests · envoydev.NET conventions for architecture fitness tests - encoding layer/dependency/naming/isolation rules as tests that fail the build on a violation, so the structure dotnet-architecture prescribes cannot erode silently. Floors at .NET 8 / C# 12. Load when adding or reviewing architecture / fitness tests or enforcing layer/dependency rules, or when the user names NetArchTest, ArchUnitNET, fitness function, architecture test, or dependency rule. Companions: dotnet-testing (the test-suite host), dotnet-architecture (the boundaries enforced), dotnet-code-quality (the analyzer/style counterpart). Do NOT load for runtime behavior tests (dotnet-testing) or analyzer/formatter config (dotnet-code-quality).
- ▌ Dotnet Web Error Handling · envoydevASP.NET Core error-handling conventions - keep the two failure channels apart (Result/typed errors for expected outcomes, exceptions caught once for the genuinely unexpected), surface everything as RFC 9457 ProblemDetails, centralize one error-to-status map, install a global IExceptionHandler (.NET 8+), and validate input in an endpoint filter with FluentValidation. Floors at .NET 8 / C# 12. Load before deciding how an API reports failures, adding a global handler, shaping error bodies, or wiring request validation, or when the user says ProblemDetails, IExceptionHandler, UseExceptionHandler, Result type, or error envelope. Companions: csharp (the throw-vs-return baseline), dotnet-minimal-api, dotnet-web-backend. Do NOT load for non-HTTP code - model expected failures with the Result half in csharp instead.
- ▌ Project Agent Capabilities · envoydevThe deliberate capabilities capture: inventory what THIS project actually has installed - the slash-only orchestration skills (from .claude/skills frontmatter), the subagent seats (.claude/agents), the MCP servers (.mcp.json), the plugins (best-effort) - and generate the always-on awareness rule .claude/rules/baseline-project-agent-capabilities.md: the fixed house usage policy plus the real inventory, never an assumed stack. Re-run after an install, a stack update, or a manifest trim - the rule is regenerated wholesale. Manual, /-only. Triggers on 'capture the project capabilities', 'refresh the capabilities rule', 'what does this project have installed'. NOT for capturing architecture (project-architecture-analyzer), code style (project-code-style-analyzer), or siblings (project-related-context).
- ▌ Project Build From Scratch · envoydev bundleBuild a new application or major module from scratch - greenfield design, scaffolding, and build orchestration, before code exists. DESIGN runs IN-SESSION on this skill's opus/xhigh pinned turn: the spec becomes 2-3 reasoned architecture options and the user picks - nothing is scaffolded before that pick. Then the stack's real new-project command + baseline wiring, then the build slice by slice - through the domain seats or in-session, per the run-start ask. Not for changing an existing codebase - a feature inside a live app is `project-solve-cross-task`, a new module in an existing repo is the project-architecture-analyzer capture plus that stack's solution-designer. Triggers on build from scratch, new project, greenfield, scaffold, start a new app.
- ▌ Project Test Coverage Loop · envoydev bundleThe deliberate coverage analyze-triage-fix loop. Runs the project-test-coverage-analyzer capture, works COVERAGE.md's weak points by tier - missing tests as scoped implementer briefs, testability refactors designer-led with plan approval first, structural gaps flagged for a user decision and never auto-applied, 'no test infrastructure' as the first substantial fix - re-runs the capture to reconcile the docs, and loops until the requirement holds or the loop plateaus. Manual, /-only. Triggers on 'run the test coverage loop' or 'raise the test coverage'. NOT for a measure-only run with no fixes (/project-test-coverage-analyzer alone), a code-quality polish (project-quality-loop), or architecture weaknesses (project-architecture-quality-loop).
- ▌ Project Code Style Analyzer · envoydev bundleThe deliberate project code-style capture: fan out code-style-analyzer agents (one per detected language), merge their reports into <docs-path>/PROJECT-CODE-STYLE.md, and generate the path-scoped project-code-style rule that auto-attaches the style core whenever a matching file is touched - in the main session AND in dispatched subagents. Re-run to refresh: the same analysis, the doc reconciles in place, the rule regenerates from the fresh reports. Manual, /-only. Triggers on 'capture the project code style' or 'set up the code-style doc and rule'. NOT for architecture (project-architecture-analyzer), one language's style question (@agent-code-style-analyzer alone), or enforcing style (the per-language configs stay the enforced source).
- ▌ Project Stack Usage Analyzer · envoydevToken/tool usage audit of claude-stack skill runs in THIS project: finds EVERY session transcript with a stack-skill run (or the SESSIONS named), runs the stack's analyze-usage.js over each, and writes a per-session report (tokens, tool calls, waste, protocol check, verdict) plus the raw data for a follow-up agent - and a cross-session SUMMARY.md when several sessions are audited. Manual, /-only. Triggers on 'analyze the stack usage', 'usage report for the skill runs', 'audit all sessions for this project', 'how many tokens did the flow burn'. NOT for live session cost (claude-hud shows that), fixing the findings (route them to the owning skill), or benchmarking model choices.
- ▌ Project Architecture Analyzer · envoydev bundleThe deliberate architecture capture: on a FIRST capture dispatch architecture-analyzer per module (the cheap read-only characterizer) - an UPDATE reconciles inline over the git diff - reason over the digests IN-SESSION - declared vs enforced structure, the stack-keyed hazards, strengths and weaknesses - and write <docs-path>/architecture/ARCHITECTURE.md, <docs-path>/architecture/ASSESSMENT.md, and the generated always-on awareness rule .claude/rules/baseline-project-architecture.md. Re-run to refresh: all three artifacts reconciled in place. Deliberate-only - run via / or as project-architecture-quality-loop's ANALYZE step, never mid-build. Triggers on 'capture the architecture' or 'refresh the architecture docs'. NOT for fixing the weaknesses it finds (project-architecture-quality-loop runs this capture first), one module's characterization (@agent-architecture-analyzer), code style (project-code-style-analyzer), or test-suite/coverage weaknesses (project-test-coverage-analyzer - the assessment keeps only st
- ▌ Project CI Failure Signatures · envoydevUse when a CI pipeline or PR check goes red and you want to triage it yourself in the current chat - match the failure to a signature, make the call between a real code defect CI surfaced first and an environment / pin / config / workflow failure that never touched the code, and route it. The single-chat form of the ci-failure-diagnoser seat. Trigger on red CI, PR check failing, passes locally but fails in CI, NU1301, ERESOLVE, exit 137, workflow YAML broke, flaky pipeline. Not a crash on your own machine (that is project-runtime-failure-signatures), not authoring CI/CD (that is devops).
- ▌ Project Test Coverage Analyzer · envoydevThe deliberate test-coverage capture: detect each stack's test-and-coverage tooling, run the instrumented suite ONCE per surface in the main session, analyze per surface (seats on a first capture, inline on an update), and judge against the USER's % bar - asked at first capture, recorded in the doc - writing <docs-path>/test-coverage/COVERAGE.md (verdicts, per-module data, weak points tiered small/substantial/structural with a simplify-testing action each) plus the raw results under raw/. Re-run to refresh: both reconciled in place. Triggers on 'measure the test coverage', 'capture the coverage baseline', or 'how covered is this project'. Manual, /-only - or Read-loaded as project-test-coverage-loop's ANALYZE step; never mid-build, never a flow gate. NOT for fixing the gaps it finds (project-test-coverage-loop), writing a task's tests (every build flow's own done bar), or architecture capture (project-architecture-analyzer).
- ▌ Project Architecture Quality Loop · envoydev bundleThe deliberate architecture analyze-assess-improve loop. Runs the project-architecture-analyzer capture, works the ASSESSMENT's weaknesses by tier - small to a domain implementer, substantial through a designer -> implementers -> verifier (plan approved first), structural flagged for a user decision and never auto-applied - prunes the fixed weaknesses from the ASSESSMENT itself (one capture reconcile at stop, not per round), and loops until the fixable cons resolve or plateau. Manual, /-only. Triggers on 'run the architecture quality loop' or 'analyze and improve the architecture'. NOT for a code-quality polish (project-quality-loop), a single feature build (project-solve-cross-task), or a capture-only run with no fixes (/project-architecture-analyzer alone).
- ▌ Project Runtime Failure Signatures · envoydevUse when something breaks at runtime on your own machine and you have the evidence - a stack trace, an exception, a hang, or a broken screen - and want to know where the real cause lives. A lookup of the common local-runtime failure signatures, each mapped to where to isolate it - usually not the line that threw. The single-chat form of the diagnoser seat's failure catalogue; pairs with the systematic-debugging method. NOT for a CI or build/test-gate failure (the resolvers and ci-failure-diagnoser own those) or a production incident - local-runtime evidence only. Keywords NullReferenceException, Cannot read properties of undefined, Unable to resolve service, NG0201 No provider, ObjectDisposedException, deadlock, hang, IndexOutOfRangeException, 401, 403, config drift.