Skills and agents
Browse 41 skills and 3 agents generated from the repository. Every card opens a standalone page that works without JavaScript; the filter below narrows the list when JavaScript is available.
- anti-slopKill generic AI patterns before they ship — filler prose, over-commenting, defensive over-engineering, unnecessary abstractions, emoji headers. Apply on any UI, copy, or code that will be read by a human, including comments and READMEs.
- api-designDesign REST or GraphQL APIs: resource modeling, URL conventions, error contracts, versioning, pagination, idempotency keys, rate-limit headers. Use when creating a new API endpoint or reviewing API consistency.
- automation-pickBefore automating a task, decide whether to automate. Decision tree based on volume, frequency, error cost, and reversibility. Use when someone says 'we should automate this' or 'let's build a script for that'.
- bottleneck-gatingPhase plans by measured bottleneck, not by requested order. Each phase has a gate condition (metric must move before proceeding). Never monetize or beautify a loop that doesn't complete. Use when planning a multi-phase project where phases haven't been ordered by measurement.
- budget-guardSet a hard spend ceiling before a paid loop starts, measure each call, and stop at the limit with a report. Covers tokens, dollars, retries, and wall clock. Use for any agent loop, batch, or retry policy that spends money.
- cache-identityUse when caching model results, latency, or artifacts: build a versioned key, bound the store, publish atomically, and never trust an unvalidated cache path.
- constraint-then-solveBefore solving, restate the problem, list unknowns, and catalog every constraint. Verify the solution against every constraint at the end. Use on any non-trivial reasoning task before jumping to a solution.
- context-optimizationKeep tool outputs small, sandbox large files, reference not repeat. Use on any task that produces bulky tool output, touches large files, or runs longer than a few minutes.
- cron-troubleshootDebug a cron job that failed or didn''t run: timezone, drift, overlap, missing logs, dependency failure. Use when a scheduled job is missing, late, or producing wrong results.
- e2e-test-strategyPlan an end-to-end test pyramid: which flows to E2E, which to integration, which to unit. Seeding strategy, auth state handling, flake mitigation, parallel execution. Use when designing test coverage for a web app or API.
- engram-routingUse an approved, hash-keyed routing instinct to annotate a repeated prompt before the planner ranks skills. Use when a prompt pattern recurs and routing should reuse prior evidence.
- execution-observabilityLog what ran, on which model, with which skill, at what cost, and whether it succeeded, without storing prompts or secrets. Query history to debug routing and spend. Use after adding execution paths or when a run must be auditable.
- fallback-model-routingRoute around provider overload, rate limits, and timeouts without hiding the switch. Retry the primary within a bounded budget, try declared fallbacks in order, and report which model answered. Use when a model call can fail and work must continue.
- feedback-loopAfter shipping, set up the loop: instrument → measure → review weekly → adjust. Use when shipping a feature, launching a product, or noticing nobody acts on data.
- judge-integrityUse when a model judges model output: separate judge failure from a low score, run deterministic checks first, gate escalation, and keep scoring separate from generation.
- knowledge-extractTurn a repeatedly observed behavior into a skill draft with a trigger, procedure, anti-patterns, and checks. Use when the same correction or workflow has appeared at least three times.
- landscape-firstBefore building anything in a competitive space, research 5-10 competitors: pricing, UX patterns, positioning, weaknesses. Extract what to steal, adapt, or avoid. Every project should do this before the first feature commit. Use when starting a feature in a competitive space.
- measure-firstBefore planning any change, query live data (DB, API, analytics, logs) to find the actual bottleneck. Static analysis reveals structure; live data reveals truth. Plans must sequence by measured bottleneck, not by requested order. Use when about to optimize, refactor, or claim a system is slow.
- plan-before-codeNo implementation without an agreed spec for non-trivial work. Brainstorm, write the spec, confirm it, then code. Use before any feature, component, behavior change, or refactor with more than one reasonable approach.
- plan-then-actPlan first, then act. State the plan in one sentence before making the first tool call, then keep each step bounded. Use when a task has multiple steps and the order matters.
- prompt-disciplineEnforce the four coding rules — think before acting, minimum viable change, surgical edits, stay on the stated goal. Use before any non-trivial change, file write, or substantive response.
- pushback-when-wrongWhen the brief contradicts verified reality, push back with evidence. Identify what the user is wrong about AND what they haven't thought of. Every plan must include a pushback section — silent execution of wrong premises wastes weeks. Use when reviewing an assertion that smells off, or an action you suspect is wrong.
- quality-protocolThe routing protocol for non-trivial work: restate, catalog constraints, choose one discipline, execute in verified increments. Use before declaring a non-trivial task complete.
- re-verify-findingsBefore executing any audit finding, claimed bug, or reported issue, re-verify it against LIVE sources (repo, DB, API, browser). Never accept stale assertions — a significant fraction of external claims are already fixed or wrong by the time you see them. Use after any prior verification claimed a finding.
- read-before-buildA plan or design document is an aspiration, not a specification. Before writing any code, inspect the actual source files. Use before implementing any plan whose claims about current code can be checked.
- session-closeoutBefore ending a session: reconcile changed files, update docs and changelog, list loose ends, write a handoff note. Apply at the end of any multi-step project.
- session-distillTurn a working session into a structured artifact covering changes, decisions, open questions, and reusable patterns. Use when wrapping up a session or handing work to another agent.
- show-your-workPublish the evidence, rejected alternatives, and open uncertainties behind a non-obvious decision without exposing private reasoning. Use when explaining a decision or defending a result.
- structured-output-contractWhen output feeds a program, define the schema before calling the model. Parse strictly, validate required fields and types, allow one repair attempt, then fail closed instead of guessing. Use when JSON, tool arguments, or any machine-read value is required.
- verify-firstEvery assertion about the codebase, product, or market is a hypothesis. Verify it against live sources before committing to action. The master discipline: verify first, act second. Apply before committing to any claim.
- workflow-decomposeDecompose any workflow into a DAG: trigger → conditions → actions → retries → observability. Mental model for n8n, Temporal, Airflow, GitHub Actions, or hand-rolled scripts. Use when designing or debugging an automation.
- agent-memory-vaultStructured, file-backed memory for autonomous agents that must recall decisions and preferences across runs. Use when a workflow needs durable facts that outlive a single session.
- agent-team-standardsLock a project's agent skill names and versions in a manifest, then report drift against that manifest. Use when a team needs reproducible skill versions across CI, cron, and developer machines.
- autonomous-briefingAssemble a dated briefing from external research and dataset scripts, labeling unavailable sections instead of inventing content. Use when you want a repeatable morning scan rather than an open-ended research session.
- book-to-skillTurn a book, PDF, or long article into a skill draft with reference chapters and a glossary. Use when converting long-form source material into a reviewable draft rather than a finished skill.
- chrome-devtools-mcp-bridgeDrive a Chrome DevTools MCP server for browser automation, network traces, console capture, and performance audits. Use when a web page must be inspected or exercised through the DevTools Protocol instead of a headless script.
- design-languageApply Apple-grade UI polish principles — restraint, coherence, intention, hierarchy through scale, negative space, motion that means. Use when generating UI, design systems, components, or marketing surfaces.
- funnel-pr-guardEvery PR that touches the front door must state which funnel step it changes and by how much. PR review discipline for landing, auth, onboarding, and signup flows. Use when reviewing a change that touches conversion-critical paths, landing, auth, onboarding, or pricing pages.
- persistent-memoryDecide when a cross-session memory backend is worth adopting, then set capture, retention, redaction, and injection limits before enabling it. Use when an agent must recall decisions across sessions and the harness has no built-in memory.
- sql-migration-trioEvery SQL migration = three files: up (forward DDL), down (reverse DDL), and schema.sql sync. Apply, drift-check, and rollback procedure. Use when writing or reviewing any SQL migration in a Supabase or Postgres project.
- token-compressionCompress large tool outputs, logs, and retrieved chunks before they crowd out context, while preserving the lines that carry decisions. Use when one tool result is large enough to dominate the working context.
- af-criticScore a completed task output on correctness, slop, and scope. The judge. Use after /af or any loop result completes.
- af-orchestratorOrchestrator agent that dispatches subtasks to specialized agents and skills.
- af-plannerDecompose a request into a skill/agent plan. Use when a request is multi-step, ambiguous, or has multiple reasonable approaches.
Legacy links of the form /catalog#skill-id redirect to the canonical page when JavaScript is available; without it, use the list above.