LucasSantana-Dev
Userstreamer-mode — a coding-agent skill for live streams and screen shares: never print secrets, env vars, tokens, IPs, or hostnames on screen; safe command alternatives; findings deferred privately.
Categories
Indexed Skills (97)
streamer-mode
Behavioral safety mode for when the user is live streaming or screen sharing — never print secrets, env vars, tokens, IPs, hostnames, or terminal-sensitive data on screen; substitute safe command alternatives; defer security findings to a private file instead of announcing them on stream. Use when the user says "streamer mode", "I'm streaming", "I'm live", "going live", "screen sharing", "recording my screen", "on stream", or "presenting/demo on Zoom/Meet".
repaint
Build production-grade frontend UI that doesn't read as "AI-made" — and behaves like an elite product, not just looks like one. A self-contained pipeline — research real references, lock a register + anchor, commit OKLCH tokens + distinctive type, design the interaction (feedback, forgiving input, keyboard, recovery), make it accessible and SEO-ready (semantic HTML, landmarks, alt, metadata), build, then iterate against slop + experience + a11y/SEO audits and a live browser-screenshot loop. Grounded in the industry-standard design systems (Material 3, Apple HIG, Carbon, Polaris, Fluent, Ant, shadcn) and the psychology of interface design (Gestalt, Hick, Fitts, peak-end), with manipulative dark patterns banned. Use for any page, component, dashboard, landing page, settings, data table, onboarding, or app surface — especially when output looks generic, templated, or "AI-made", feels hollow, or is inaccessible/invisible to search; aim at Linear/Vercel/Stripe/Notion quality. Ships a reference library + 23 evals.
frontend-reference-hunt
Search and map visual frontend references from Dribbble, Mobbin, Pinterest, and curated galleries (Awwwards, Godly, Land-book, siteinspire) into a structured reference board covering register, typography, palette, spacing, layout, and named motion archetypes, with steal/avoid notes per reference. Use when a UI task needs real-world visual direction before building ("busca referências", "find design references", "reference hunt", "moodboard pra essa página", "procura no Dribbble/Mobbin", "referências de UI/UX/motion") or when a build skill needs its reference-anchor phase fed with live, current references instead of training-data memory. Searches and maps only; does not build UI.
three-man-team
Parallel Architect→Builder→Reviewer pattern for complex multi-phase tasks. Dispatches 3 agents: Architect (Opus) reads codebase and writes a plan with phases; Builder (Sonnet) implements from plan and stages changes; Reviewer (Sonnet) validates diff and runs tests. Cuts turnaround on 3+ phase features and big PRs from 6+ turns to 1-turn dispatch. Use when phases justify the overhead; when CRITICAL-flagged; or for large PRs touching >5 files.
backlog
Composite backlog builder for a single repo. Analyzes code, scores findings by value and ROI, groups into themes, and creates prioritized GitHub issues on a Project board. Use to discover what to work on, build a backlog, plan sprints, or prioritize work with a time budget.
decide
Makes an explicit decision with documented alternatives and rationale. Composite skill — two-phase pipeline that chains research-and-decide (research options + recommendation) → adr-write (record at proportional fidelity: DECISIONS.md line by default, full ADR only behind the record gate). Stops after Phase 1 if inconclusive. Use when making architectural or tooling decisions that need both recommendation and a durable record.
dep-sweep
Batch-process Dependabot/Renovate PRs by risk: auto-merge safe ones (devDeps, patches) into the configured base branch (main by default), surface risky ones for human review. Chains PR enumeration, risk classification, merge-readiness checks, and changelog batching. Use when bot PRs pile up; reduces a 20-PR queue to actionable groups.
dispatch
Parallelize independent evidence-gathering or audit work into ≥2 concurrent agent tracks; reconcile into a single verdict and action. Also dispatches named agent-box task scripts or ad-hoc prompts to remote <homelab> infrastructure over SSH. Use when sub-tasks are cleanly separable with no output dependencies and running serially wastes turns. Examples: multi-repo scans, concurrent CI/lint/test audits, N-package dependency sweeps. Skip when tasks depend on prior output.
efficiency-advisor
Analyze a proposed multi-agent workflow, Workflow() script, or active session for token waste and time bottlenecks. Catches model-tier mismatches, sequential work that should be parallel, re-read waste, and context bloat — scores each fix by impact with estimated savings. Use before spawning large agent fleets, when choosing between parallel vs sequential dispatch, or when a session is burning budget faster than expected.
knowledge-loop
Composite skill — query, capture, improve, and persist knowledge in one workflow. Chains recall (RAG query) → sync-memories (write durable note) → rag-curate (improve weak retrievals) → handoff (durable snapshot if session-ending). Use when the work involves "what did we decide", "remember this", "save where we are", or any closing checkpoint.
orchestrate
Coordinate multi-agent teams, multi-step or multi-repo work across plans, skills, worktrees, and parallel investigations. Use when the task is large enough to benefit from parallel work, independent workstreams, and one lead agent owning synthesis.
pr-merge-readiness
Aggregate every signal that gates a PR merge (CI, reviews, conflicts, branch staleness, security scans, third-party reviewers like CodeRabbit/Greptile/Sonar) into a single readiness verdict — MERGE / WAIT / FIX. Use as the one-call check before clicking merge instead of running ci-watch + gh-fix-ci + manual scan.
research-and-decide
Composite skill — research a question, challenge the conclusion, plan adoption, and capture the decision. Chains adt-research or brainstorming (explore) → critic agent (challenge) → plan (sequence) → adr-write (record). Use when evaluating a library / pattern / architecture choice — forces the research-to-ADR pairing that usually slips.
session-bootstrap
Composite skill — start-of-day routine. Chains wake-up (compact brief) → next-priority (decide) → pr-snapshot (PR queue) → context-pack (load relevant context if work-intent). Replaces "what was I doing", "what's next", "load context" with one invocation. Auto-fires on the first non-trivial prompt of a fresh session.
streamer-mode
Behavioral safety mode for when the user is live streaming or screen sharing — never print secrets, env vars, tokens, IPs, hostnames, or terminal-sensitive data on screen; substitute safe command alternatives; defer security findings to a private file instead of announcing them on stream. Use when the user says "streamer mode", "I'm streaming", "I'm live", "going live", "screen sharing", "recording my screen", "on stream", or "presenting/demo on Zoom/Meet".
sync-memories
Sync durable project or session knowledge into the available memory systems so future sessions have accurate context. Use when meaningful work is complete and the user wants the result remembered.
sync-sharekit-profile
Sync ~/.claude/ (skills, hooks, standards, CLAUDE.md, agents) into the public sharekit profile repo, sanitize personal paths/identity, scan for secrets, show diff, and commit+push. Invoke whenever: "sync my sharekit profile", "update sharekit", "push my latest skills to sharekit", "sharekit is out of date", or after a session that significantly expanded the skill or agent library.
test-driven-development
Use when implementing any feature or bugfix, before writing implementation code. Also invoked via /tdd (alias archived — this is the canonical skill).
version-bump
Automate version bumping in npm monorepos with CHANGELOG promotion and PR automation
backlog
Composite backlog builder for a single repo. Analyzes code, scores findings by value and ROI, groups into themes, and creates prioritized GitHub issues on a Project board. Use to discover what to work on, build a backlog, plan sprints, or prioritize work with a time budget.
decide
Makes an explicit decision with documented alternatives and rationale. Composite skill — two-phase pipeline that chains research-and-decide (research options + recommendation) → adr-write (record at proportional fidelity: DECISIONS.md line by default, full ADR only behind the record gate). Stops after Phase 1 if inconclusive. Use when making architectural or tooling decisions that need both recommendation and a durable record.
dep-sweep
Batch-process Dependabot/Renovate PRs by risk: auto-merge safe ones (devDeps, patches) into the configured base branch (main by default), surface risky ones for human review. Chains PR enumeration, risk classification, merge-readiness checks, and changelog batching. Use when bot PRs pile up; reduces a 20-PR queue to actionable groups.
dispatch
Parallelize independent evidence-gathering or audit work into ≥2 concurrent agent tracks; reconcile into a single verdict and action. Also dispatches named agent-box task scripts or ad-hoc prompts to remote <homelab> infrastructure over SSH. Use when sub-tasks are cleanly separable with no output dependencies and running serially wastes turns. Examples: multi-repo scans, concurrent CI/lint/test audits, N-package dependency sweeps. Skip when tasks depend on prior output.
efficiency-advisor
Analyze a proposed multi-agent workflow, Workflow() script, or active session for token waste and time bottlenecks. Catches model-tier mismatches, sequential work that should be parallel, re-read waste, and context bloat — scores each fix by impact with estimated savings. Use before spawning large agent fleets, when choosing between parallel vs sequential dispatch, or when a session is burning budget faster than expected.
knowledge-loop
Composite skill — query, capture, improve, and persist knowledge in one workflow. Chains recall (RAG query) → sync-memories (write durable note) → rag-curate (improve weak retrievals) → handoff (durable snapshot if session-ending). Use when the work involves "what did we decide", "remember this", "save where we are", or any closing checkpoint.
orchestrate
Coordinate multi-agent teams, multi-step or multi-repo work across plans, skills, worktrees, and parallel investigations. Use when the task is large enough to benefit from parallel work, independent workstreams, and one lead agent owning synthesis.
pr-merge-readiness
Aggregate every signal that gates a PR merge (CI, reviews, conflicts, branch staleness, security scans, third-party reviewers like CodeRabbit/Greptile/Sonar) into a single readiness verdict — MERGE / WAIT / FIX. Use as the one-call check before clicking merge instead of running ci-watch + gh-fix-ci + manual scan.
research-and-decide
Composite skill — research a question, challenge the conclusion, plan adoption, and capture the decision. Chains adt-research or brainstorming (explore) → critic agent (challenge) → plan (sequence) → adr-write (record). Use when evaluating a library / pattern / architecture choice — forces the research-to-ADR pairing that usually slips.
session-bootstrap
Composite skill — start-of-day routine. Chains wake-up (compact brief) → next-priority (decide) → pr-snapshot (PR queue) → context-pack (load relevant context if work-intent). Replaces "what was I doing", "what's next", "load context" with one invocation. Auto-fires on the first non-trivial prompt of a fresh session.
streamer-mode
Behavioral safety mode for when the user is live streaming or screen sharing — never print secrets, env vars, tokens, IPs, hostnames, or terminal-sensitive data on screen; substitute safe command alternatives; defer security findings to a private file instead of announcing them on stream. Use when the user says "streamer mode", "I'm streaming", "I'm live", "going live", "screen sharing", "recording my screen", "on stream", or "presenting/demo on Zoom/Meet".
sync-memories
Sync durable project or session knowledge into the available memory systems so future sessions have accurate context. Use when meaningful work is complete and the user wants the result remembered.
sync-sharekit-profile
Sync ~/.claude/ (skills, hooks, standards, CLAUDE.md, agents) into the public sharekit profile repo, sanitize personal paths/identity, scan for secrets, show diff, and commit+push. Invoke whenever: "sync my sharekit profile", "update sharekit", "push my latest skills to sharekit", "sharekit is out of date", or after a session that significantly expanded the skill or agent library.
test-driven-development
Use when implementing any feature or bugfix, before writing implementation code. Also invoked via /tdd (alias archived — this is the canonical skill).
version-bump
Automate version bumping in npm monorepos with CHANGELOG promotion and PR automation
catalog-gardener
Audit skill catalog for rot, dead links, oversize files, and orphans. Detect stale generated artifacts, skills exceeding size limits, and skills not referenced by any composite or routing rule. Output severity-ranked report with remediation hints.
phase-0-audit
Pre-build audit with dedup checks before creating new skills or agents. Check existing agents/skills before creating new ones. Change-type→phases matrix: full containment→reuse, partial+generalizable→extend, intentional domain-specific→new, completely different→new. Dedup checks before skill/agent creation.
three-layer-eval
Three-layer skill quality evaluation (Static → LLM Judge → Monte Carlo). Layer 1: deterministic checks (frontmatter, triggers, size, links) in <2s. Layer 2: semantic eval across 4 dimensions using Haiku+Sonnet. Layer 3: 50-100 simulated runs for statistical reliability.
adt-rag-drift
Detect and fix stale chunks (files that changed or were deleted since last indexing)
app-teardown
Reverse-engineer a live application (web app, SaaS, mobile) by walking its real flows in the browser and map every idea worth taking — UX patterns, features, visual design, copy, growth mechanics — into a teardown report where each finding carries evidence, an adopt/adapt/already-have/reject verdict, and a constraint rationale for our context. Use when asked "faz um teardown do app X", "analisa esse app e mapeia ideias", "what can we steal from X", "reverse-engineer this product", "competitive teardown", "o que dá pra aproveitar do app Y". For a codebase or repository use code-teardown; for purely visual reference boards from design platforms use frontend-reference-hunt.
changelog-update
Updates CHANGELOG.md from recent commits, grouping entries by type (feat/fix/docs/breaking) with semantic version headings. Update CHANGELOG.md by promoting [Unreleased] content to a versioned section and bumping the package version. Follows Keep a Changelog format. Use when preparing a release or when [Unreleased] has accumulated work that needs to be captured under a version.
ci-watch
Diagnose failing CI checks, isolate the first real blocker (vs. noise), and surface the smallest viable fix. Use when a PR has failing checks, the pipeline is broken, or you need to understand why CI is red before proceeding with merge or release.
code-teardown
Reverse-engineer an external repository or codebase (competitor OSS, reference implementation, trending project) and map every adoptable idea — architecture, code patterns, DX, tooling, testing, docs conventions — into a teardown report where each finding carries evidence (file:line), an adopt/adapt/already-have/reject verdict, and a constraint rationale for our context. Use when asked "faz um teardown desse repo", "analisa esse repositório pra ideias", "o que dá pra aproveitar desse projeto open-source", "evaluate X repo for insights", "what can we learn from this codebase". Read-only toward the target; clones to External HD. For live apps use app-teardown; to onboard for contributing use adt-repo-intake; to document our own architecture use doc-and-modernize.
codebase-memory
Use the codebase knowledge graph for structural code queries. Triggers on: explore the codebase, understand the architecture, what functions exist, show me the structure, who calls this function, what does X call, trace the call chain, find callers of, show dependencies, impact analysis, dead code, unused functions, high fan-out, refactor candidates, code quality audit, graph query syntax, Cypher query examples, edge types, how to use search_graph.
decide-now
Force a decision when the agent is stuck in analysis paralysis, going in circles, or repeatedly weighing options without committing. Picks the best available option, states the rationale, and moves forward immediately. Use when the agent has been deliberating for 3+ turns, when options are deadlocked, or when the user says "just decide", "pick one", "move on", or "stop deliberating". Also use when the user asks for a decision and the agent keeps hedging.
frontend-reference-hunt
Search and map visual frontend references from Dribbble, Mobbin, Pinterest, and curated galleries (Awwwards, Godly, Land-book, siteinspire) into a structured reference board covering register, typography, palette, spacing, layout, and named motion archetypes, with steal/avoid notes per reference. Use when a UI task needs real-world visual direction before building ("busca referências", "find design references", "reference hunt", "moodboard pra essa página", "procura no Dribbble/Mobbin", "referências de UI/UX/motion") or when repaint or frontend-craft needs its reference-anchor phase fed with live, current references instead of training-data memory. Searches and maps only; does not build UI.
graphify
any input (code, docs, papers, images, videos) to knowledge graph. Use when user asks any question about a codebase, documents, or project content - especially if graphify-out/ exists, treat the question as a /graphify query.
loop-engineer
Designs and implements autonomous agent loops for any task. Trigger when the user wants to stop manually prompting agents and build a self-running feedback cycle — coding automation, research pipelines, content workflows, CI loops, or any repeating AI-driven process. Also trigger for "design a loop", "build an agent loop", "I want this to run automatically", or when describing a workflow currently driven manually. Output: complete loop spec plus working implementation.
memory-promote
Promote a personal-scope session learning into team scope through curated review. Strips private-tagged content and secret-shaped strings, then produces a reviewable artifact (PR-style proposal) instead of writing shared memory directly. Use when a personal memory note is worth sharing with the team. Enforces .harness/memory-scopes.json promotion rules.
memory-prune
Audit project memory files for stale entries — when cited PRs merge, bugs fix, or files disappear. Proposes archival or deletion per entry. Read-only by default; prompts before changing. Use periodically (monthly) or when memory recall contradicts current code. Pairs with sync-memories (fresh notes) and adt-rag-drift (RAG index handling).
next-priority
Decide the highest-value safe thing to do right now in the active repo or workspace.
parallel-work-coordinator
Orchestrate 3–5 independent tasks in parallel with auto-dispatch and reconciliation. Use for multiple independent work items (multi-repo audits, translations, investigations, sweeps). Invoke on "do all these", "check all repos", "audit in parallel", or "handle independently". Wraps multi-agent dispatch for lightweight jobs; use Workflow for heavy orchestration (loops, conditionals, >5 units).
plan
Create a structured implementation plan before starting complex work. Breaks work into small phases, each with exact file paths, a runnable verification command, a checkable done-condition, and replanning triggers, plus explicit goal and in/out-of-scope boundaries. Use when the user says "plan this", "create a plan", "break this down", "phase this work", or when a task has multiple phases or unknowns. Also use when the agent needs to think before acting on complex tasks.
prototype
Build a throwaway prototype to flesh out a design before committing to it. Routes between two branches — a runnable terminal app for state/business-logic questions, or several radically different UI variations toggleable from one route. Use when the user wants to prototype, sanity-check a data model or state machine, mock up a UI, explore design options, or says "prototype this", "let me play with it", "try a few designs".
rag-curate
Improve RAG corpus quality by adding missing docs, rewriting weak chunks, and filling retrieval gaps. Use after a diagnostic skill (rag-quality, adt-rag-coverage, adt-rag-drift) flags a coverage gap or when the index returns poor/irrelevant results for a known query.
recall
One-shot semantic lookup against the local RAG index — answers "what did we decide about X", "where did we hit this bug before", "is there a memory note for Y" in a single MCP call. Backed by the rag_query MCP tool over ~21k chunks (memory, plans, handoffs, skills, standards, repo docs, changelogs, specs, code, commits, transcripts) across curated repos. Auto-scopes to current repo. Use instead of grep when the question is fuzzy, cross-file, or about prior reasoning. Skip for pure navigation.
refactor
Surgical code refactoring to improve maintainability without changing behavior. Invoke when user says "refactor this", "clean up this code", "extract functions", "this is messy", "improve maintainability", "fix code smells", "simplify this", "this function is too long", or when you see god functions, nested conditionals, duplicate logic, magic numbers, unclear naming, or dead code. Always run a discovery pass first — catalog smells before touching anything. Gradual evolution, not revolution.
review
Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes — Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/PRD asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".
secure
Run a security-first pass for config, credentials, dependency risk, unsafe code patterns, and unsafe operational shortcuts. Covers OWASP Top 10, secure coding, infrastructure config (Terraform/K8s/Docker/GitHub Actions), HTTPS/CORS/CSRF/rate limiting, and credential handling.
xp
Extreme Programming for AI-human pair development. Use when: pair programming with your AI, building incrementally with tests, working with clear role division, or the user mentions XP/YAGNI/simple design/continuous refactoring.
adt-rag-drift
Detect and fix stale chunks (files that changed or were deleted since last indexing)
app-teardown
Reverse-engineer a live application (web app, SaaS, mobile) by walking its real flows in the browser and map every idea worth taking — UX patterns, features, visual design, copy, growth mechanics — into a teardown report where each finding carries evidence, an adopt/adapt/already-have/reject verdict, and a constraint rationale for our context. Use when asked "faz um teardown do app X", "analisa esse app e mapeia ideias", "what can we steal from X", "reverse-engineer this product", "competitive teardown", "o que dá pra aproveitar do app Y". For a codebase or repository use code-teardown; for purely visual reference boards from design platforms use frontend-reference-hunt.
changelog-update
Updates CHANGELOG.md from recent commits, grouping entries by type (feat/fix/docs/breaking) with semantic version headings. Update CHANGELOG.md by promoting [Unreleased] content to a versioned section and bumping the package version. Follows Keep a Changelog format. Use when preparing a release or when [Unreleased] has accumulated work that needs to be captured under a version.
ci-watch
Diagnose failing CI checks, isolate the first real blocker (vs. noise), and surface the smallest viable fix. Use when a PR has failing checks, the pipeline is broken, or you need to understand why CI is red before proceeding with merge or release.
code-teardown
Reverse-engineer an external repository or codebase (competitor OSS, reference implementation, trending project) and map every adoptable idea — architecture, code patterns, DX, tooling, testing, docs conventions — into a teardown report where each finding carries evidence (file:line), an adopt/adapt/already-have/reject verdict, and a constraint rationale for our context. Use when asked "faz um teardown desse repo", "analisa esse repositório pra ideias", "o que dá pra aproveitar desse projeto open-source", "evaluate X repo for insights", "what can we learn from this codebase". Read-only toward the target; clones to External HD. For live apps use app-teardown; to onboard for contributing use adt-repo-intake; to document our own architecture use doc-and-modernize.
codebase-memory
Use the codebase knowledge graph for structural code queries. Triggers on: explore the codebase, understand the architecture, what functions exist, show me the structure, who calls this function, what does X call, trace the call chain, find callers of, show dependencies, impact analysis, dead code, unused functions, high fan-out, refactor candidates, code quality audit, graph query syntax, Cypher query examples, edge types, how to use search_graph.
decide-now
Force a decision when the agent is stuck in analysis paralysis, going in circles, or repeatedly weighing options without committing. Picks the best available option, states the rationale, and moves forward immediately. Use when the agent has been deliberating for 3+ turns, when options are deadlocked, or when the user says "just decide", "pick one", "move on", or "stop deliberating". Also use when the user asks for a decision and the agent keeps hedging.
frontend-reference-hunt
Search and map visual frontend references from Dribbble, Mobbin, Pinterest, and curated galleries (Awwwards, Godly, Land-book, siteinspire) into a structured reference board covering register, typography, palette, spacing, layout, and named motion archetypes, with steal/avoid notes per reference. Use when a UI task needs real-world visual direction before building ("busca referências", "find design references", "reference hunt", "moodboard pra essa página", "procura no Dribbble/Mobbin", "referências de UI/UX/motion") or when repaint or frontend-craft needs its reference-anchor phase fed with live, current references instead of training-data memory. Searches and maps only; does not build UI.
graphify
any input (code, docs, papers, images, videos) to knowledge graph. Use when user asks any question about a codebase, documents, or project content - especially if graphify-out/ exists, treat the question as a /graphify query.
loop-engineer
Designs and implements autonomous agent loops for any task. Trigger when the user wants to stop manually prompting agents and build a self-running feedback cycle — coding automation, research pipelines, content workflows, CI loops, or any repeating AI-driven process. Also trigger for "design a loop", "build an agent loop", "I want this to run automatically", or when describing a workflow currently driven manually. Output: complete loop spec plus working implementation.
memory-promote
Promote a personal-scope session learning into team scope through curated review. Strips private-tagged content and secret-shaped strings, then produces a reviewable artifact (PR-style proposal) instead of writing shared memory directly. Use when a personal memory note is worth sharing with the team. Enforces .harness/memory-scopes.json promotion rules.
memory-prune
Audit project memory files for stale entries — when cited PRs merge, bugs fix, or files disappear. Proposes archival or deletion per entry. Read-only by default; prompts before changing. Use periodically (monthly) or when memory recall contradicts current code. Pairs with sync-memories (fresh notes) and adt-rag-drift (RAG index handling).
next-priority
Decide the highest-value safe thing to do right now in the active repo or workspace.
parallel-work-coordinator
Orchestrate 3–5 independent tasks in parallel with auto-dispatch and reconciliation. Use for multiple independent work items (multi-repo audits, translations, investigations, sweeps). Invoke on "do all these", "check all repos", "audit in parallel", or "handle independently". Wraps multi-agent dispatch for lightweight jobs; use Workflow for heavy orchestration (loops, conditionals, >5 units).
plan
Create a structured implementation plan before starting complex work. Breaks work into small phases, each with exact file paths, a runnable verification command, a checkable done-condition, and replanning triggers, plus explicit goal and in/out-of-scope boundaries. Use when the user says "plan this", "create a plan", "break this down", "phase this work", or when a task has multiple phases or unknowns. Also use when the agent needs to think before acting on complex tasks.
prototype
Build a throwaway prototype to flesh out a design before committing to it. Routes between two branches — a runnable terminal app for state/business-logic questions, or several radically different UI variations toggleable from one route. Use when the user wants to prototype, sanity-check a data model or state machine, mock up a UI, explore design options, or says "prototype this", "let me play with it", "try a few designs".
rag-curate
Improve RAG corpus quality by adding missing docs, rewriting weak chunks, and filling retrieval gaps. Use after a diagnostic skill (rag-quality, adt-rag-coverage, adt-rag-drift) flags a coverage gap or when the index returns poor/irrelevant results for a known query.
recall
One-shot semantic lookup against the local RAG index — answers "what did we decide about X", "where did we hit this bug before", "is there a memory note for Y" in a single MCP call. Backed by the rag_query MCP tool over ~21k chunks (memory, plans, handoffs, skills, standards, repo docs, changelogs, specs, code, commits, transcripts) across curated repos. Auto-scopes to current repo. Use instead of grep when the question is fuzzy, cross-file, or about prior reasoning. Skip for pure navigation.
refactor
Surgical code refactoring to improve maintainability without changing behavior. Invoke when user says "refactor this", "clean up this code", "extract functions", "this is messy", "improve maintainability", "fix code smells", "simplify this", "this function is too long", or when you see god functions, nested conditionals, duplicate logic, magic numbers, unclear naming, or dead code. Always run a discovery pass first — catalog smells before touching anything. Gradual evolution, not revolution.
review
Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes — Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/PRD asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".
secure
Run a security-first pass for config, credentials, dependency risk, unsafe code patterns, and unsafe operational shortcuts. Covers OWASP Top 10, secure coding, infrastructure config (Terraform/K8s/Docker/GitHub Actions), HTTPS/CORS/CSRF/rate limiting, and credential handling.
xp
Extreme Programming for AI-human pair development. Use when: pair programming with your AI, building incrementally with tests, working with clear role division, or the user mentions XP/YAGNI/simple design/continuous refactoring.
brainstorming
Facilitate collaborative idea exploration and turn rough concepts into a validated design direction before implementation starts. Use when the user is still shaping the solution and should not be pushed into code yet.
context-pack
Packs the most relevant context (files, docs, prior decisions) into a compact payload for injection into the next agent turn. Build a task-aware context bundle before large changes, reviews, or unfamiliar work. Pul l only the code, standards, plans, RAG hits, and notes that matter instead of reading wi de. Use before refactors, cross-file fixes, unfamiliar repos, or when starting a session where you will spend more than five reads exploring. Backed by the local RAG index via the rag_query MCP tool. Skip when the task touches one known file or the answer fits in a single grep.
handoff
Compact the current conversation into a handoff document for another agent to pick up.
loop
Detect when the agent is repeating the same action or approach without progress, then break the cycle. Identifies loop patterns: trying the same fix repeatedly, re-reading the same files, re-running the same commands, or circling between alternatives. Injects a pattern interrupt, suggests alternative approaches, and forces a different strategy. Use when the agent seems stuck in a loop, when the user says "you're going in circles", or when the same error recurs after 3+ attempts.
resume
Rehydrate the current task from handoffs, plans, tasks, memory notes, and git state.
ship
End-to-end shipping — version bump, changelog update, tag, GitHub release, deploy, and post-deploy verification. Validates PR goal, checks, review state, and risk. Refuses admin bypasses or force options against main.
tdd
Short alias for test-driven development. Use for red-green-refactor, tracer bullets, behavior-driven testing, or TDD workflows.
verify
Run the narrowest meaningful validation sequence before merge, release, or handoff.
brainstorming
Facilitate collaborative idea exploration and turn rough concepts into a validated design direction before implementation starts. Use when the user is still shaping the solution and should not be pushed into code yet.
context-pack
Packs the most relevant context (files, docs, prior decisions) into a compact payload for injection into the next agent turn. Build a task-aware context bundle before large changes, reviews, or unfamiliar work. Pul l only the code, standards, plans, RAG hits, and notes that matter instead of reading wi de. Use before refactors, cross-file fixes, unfamiliar repos, or when starting a session where you will spend more than five reads exploring. Backed by the local RAG index via the rag_query MCP tool. Skip when the task touches one known file or the answer fits in a single grep.
handoff
Compact the current conversation into a handoff document for another agent to pick up.
loop
Detect when the agent is repeating the same action or approach without progress, then break the cycle. Identifies loop patterns: trying the same fix repeatedly, re-reading the same files, re-running the same commands, or circling between alternatives. Injects a pattern interrupt, suggests alternative approaches, and forces a different strategy. Use when the agent seems stuck in a loop, when the user says "you're going in circles", or when the same error recurs after 3+ attempts.
resume
Rehydrate the current task from handoffs, plans, tasks, memory notes, and git state.
ship
End-to-end shipping — version bump, changelog update, tag, GitHub release, deploy, and post-deploy verification. Validates PR goal, checks, review state, and risk. Refuses admin bypasses or force options against main.
tdd
Short alias for test-driven development. Use for red-green-refactor, tracer bullets, behavior-driven testing, or TDD workflows.
verify
Run the narrowest meaningful validation sequence before merge, release, or handoff.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.