Skills
Generated reference pages for every SKILL.md shipped by the marketplace.
adk-core
auto- Top-level prompt-routing dispatcher for the adk marketplace. The default entry point for any non-trivial request when the user has not named a specific adk-* skill. Expands the prompt, resolves entities against…config-update- Refresh ~/.config/adk/*.md against live sources to keep drift-prone fields current. Reads each existing meta-info file, queries the topic’s source-of-truth (Datadog dashboards/monitors/services, Statsig active…context-gather- Follow links in a prompt or input file (Jira, Confluence, Google Doc, Slack message/thread, Gmail thread, GitHub PR/issue) and produce a single deduplicated context.md summarizing each source. Uses the claude.ai…info- Read and merge ~/.config/adk/*.md into a single structured JSON view that other skills consume. Use when an outer skill needs current operator/repo/MCP context, when debugging “why did the skill pick service X?”, or…mode-contract- The universal definition of--auto,-i,--fix, and--auto --fixmodes used by every adk skill. Reference-only skill: it documents the contract and provides the parsing helpers other skills source. Use when…prompt-expand- Standalone prompt-expansion skill. Takes a free-form prompt and outputs a structured plan: restated intent, resolved entities, classified verb, link list, recommended skill chain. Use when an outer skill needs to…setup- Bootstrap a fresh adk install on a machine. Creates ~/.config/adk/*.md from templates (info, repos, github, datadog, mixpanel, statsig, snowflake, slack, review, docs); checks CLI deps (gh, jq, fd, ripgrep, fzf,…temp-folder- Canonical .temp/task-/ working-artifact layout. Every other adk skill writes through this convention. Use this skill whenever you need to create a new task workspace, look up the path for a particular artifact,…
adk-code
code-api- Design or evolve a stable interface — REST/HTTP API, RPC contract, internal module boundary, library export surface, CLI flag set, public type definitions — using contract-first discipline, Hyrum’s Law awareness,…code-bugfix- Fix a bug with explicit root-cause analysis, the smallest correct patch, and a regression test that fails on the buggy commit and passes on the fix. Different from/adk-code:code-write: a bugfix MUST start with a…code-migrate- Migrate a framework, runtime, library version, or build tool with explicit breaking-change analysis, scoped per-file changes, and validation between groups. Reads the upstream migration guide via WebFetch,…code-perf- Diagnose and fix a performance regression or hit a stated performance budget — measure first, identify the bottleneck with quoted profiler / trace evidence, apply the smallest correct fix, verify the win with…code-refactor- Restructure code WITHOUT changing observable behavior — rename, extract, inline, dedupe, simplify, reshape modules, split a long file — while keeping tests green throughout. Each micro-step leaves the suite green;…code-security- Implement a security-hardening change — input validation at the boundary, authn/authz fix, secrets handling, security header (CSP/HSTS/etc.), CORS policy, dependency CVE patch, rate limit, CSRF/XSS/SQL-injection…code-test- Author or expand automated tests — unit, integration, end-to-end — to cover behavior, raise coverage on a target module, lock in a previously-shipped fix, or convert manual checks into repeatable validation. Each…code-write- Implement a new feature or general code change in an existing codebase. Reads the repo’s conventions (AGENTS.md / CLAUDE.md / .cursorrules / recent commits) BEFORE writing, plans the smallest correct change in…
adk-review
audit-pr- Fast fixed-set audit on a single PR diff (NOT a deep review). Runs a parallel checklist: lint clean on changed files, typecheck clean, tests-added vs LOC-added heuristic, secrets-in-diff, license headers on new…audit-repo- Whole-repo multi-dimensional audit — security, performance, code quality, dependencies, test coverage, architecture — producing a single severity-tiered report with file-anchored evidence per finding. Read-only:…review-code-changes- Self-review of LOCAL working-tree changes (uncommitted, staged, or branch-vs-baseline) — before pushing or opening a PR. Picks up the baseline automatically: the branch’s tracking branch, then…review-feedback- Triage existing reviewer comments on a GitHub PR and address each in code with traceable per-comment replies (commit SHA + one-line summary). Different fromreview-pr --fix: this skill ASSUMES the comments are…review-handoff- Capture a structured session-handoff document so a paused or transferred task can resume without information loss. Reads the canonical .temp/task-/ artifact set, the current git state (branch, dirty files, last… review-pr- Review a remote PR (your own or a peer’s) on GitHub. Performs a fresh full-scope review every run (not just delta-since-last-review), validates existing inline comments / replies / resolved tasks, drafts new findings…
adk-docs
docs-changelog- Append or update a CHANGELOG.md entry from a set of commits or a release-tag range. Readsgit log <from>..<to>and groups commits by type (Added / Changed / Deprecated / Removed / Fixed / Security per Keep a…docs-commit-message- Draft a commit message from the staged diff (git diff --cached). Detects the repo’s existing convention — Conventional Commits, semantic-release format, or free-form — fromgit log -10 --pretty=format:%sand…docs-diagram- Author a Mermaid diagram (.mermaid / .mmd) from a description OR from real code / config evidence. Supports flowchart, sequence, class, state, ER, gantt, gitgraph, mindmap, timeline, and C4. Keeps every diagram under…docs-pr-description- Draft a PR description from the actual diff and commit history — NOT from the branch name alone. Readsgit log <base>..HEAD,git diff <base>...HEAD, and any linked ticket context; loads…docs-publish-confluence- Publish an existing markdown document as a Confluence page (create or update) via the workspace Atlassian connector — NO MCP ships with this plugin. Supports parent-page placement, label management, and post-publish…docs-publish-gdrive- Publish a markdown document to Google Drive as a Google Doc, plain markdown file, or PDF via the workspace Google Drive connector (NO MCP ships with this plugin). Supports folder placement, format selection, and…docs-review- Review an existing technical document — a markdown file in the repo, a fetched URL, a Confluence page via the workspace Atlassian connector, or a Google Doc via the workspace Google Drive connector — for accuracy…docs-write- Author or refresh a technical markdown document — README, runbook, ADR, API reference, onboarding guide, migration guide, design doc — grounded in the actual code / configs / commits of the repo (not training data)…
adk-investigate
investigate-datadog- Query Datadog (logs, metrics, monitors, traces, dashboards, error tracking, incidents) via the hosted Datadog MCP server. Three modes-of-use:investigate(free-form question to targeted query + summary),…investigate-deploy- Recent deploys timeline for one or more repos viagh run list(CLI). Surfaces deploy workflow runs in window, status, duration, triggering commit, author, and workflow URL. Use when correlating prod symptoms to…investigate-experiment- Cross-check a Statsig experiment’s pulse against Mixpanel reality and Datadog guardrails. Statsig says “primary metric is up 5%”; this skill validates by pulling the same metric from Mixpanel (project-level, not…investigate-incident- Full incident-investigation workflow combining Datadog (logs + metrics + traces + monitor history), recent deploy timeline (viaghCLI), and (optionally) a Slack channel scrape via the workspace connector. Accepts…investigate-mixpanel- Query Mixpanel for product-analytics insights via the workspace Mixpanel connector. Three modes-of-use:usage-summary(top events, DAU / WAU / retention),funnel(conversion through a sequence of events),…investigate-rca- Full root-cause-analysis composite. Runs/adk-investigate:investigate-incident(DD logs/metrics/traces/monitors + recent deploys + Slack — accepts a free-form symptom OR URL inputs like a Slack alert permalink, DD…investigate-snowflake- Read non-PII data from Snowflake via the workspace QDP_SNOWFLAKE_MCP_SERVER connector. Surfaces the connection-path decision (local MCP / prod service account / MCP wired to a prod app — same pattern as…investigate-statsig- Query Statsig for experiment results, gate evaluations, audit logs, and metric definitions via the hosted Statsig MCP. Five modes-of-use:pulse(read experiment results — primary metric, secondary metrics,…