agents-skills

Plugin agents-skills from marketplace agents-skills in skyfox675/agents-skills. Commands read from the repo's manifest — yesterday.

Skills (32)

• browser-diagnosis

  Reproduce and root-cause a front-end bug in a real browser using whichever browser MCP is connected — Chrome (claude-in-chrome), Playwright, Cypress, or Chrome DevTools — not just one. Drive the page as the signed-in user, watch the network and console, capture the failing request/response, shim/moc

• caveman

  Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says \"caveman mode\", \"talk like caveman\", \"use caveman\", \"le

• ci-flake-hunting

  The authoritative method for hunting, root-causing, and fixing CI flakes so the merge queue stops ejecting correct PRs and churn drops. A flake is a test that fails then passes on retry with NO code change explaining it — every red speculative merge-queue build ejects a correct PR and disarms its au

• ci-speed-hunting

  The authoritative method for hunting, diagnosing, and fixing CI SLOWNESS so the merge queue drains faster and throughput rises — the target is wall-clock latency, NOT correctness (a job that is GREEN but slow). Encodes the loop — MINE CI timing for the real long poles → DIAGNOSE the bottleneck class

• deploy-watching

  Single-lane interval watcher that keeps the integration environment's deploys healthy by fixing FORWARD — the deploy-health lane of a multi-agent fleet. Each tick it reads the newest deploy run, locates the red (deploy step vs post-deploy smoke), classifies it transient (concurrent-deploy state-lock

• dispatching-subagents

  Turn a groomed tracker issue into a running implementation sub-agent and a merged PR — issue selection (priority, file-disjointness, bundling), model/effort resolution from chat > labels > defaults, the eight-section dispatch brief (worktree, toolchain prefix, pre-push hygiene, scope, git discipline

• doc-coauthoring

  Guide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc

• docx

  Use this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterhea

• driving-prs-to-merge

  Owns the entire post-open PR lifecycle in a multi-agent GitHub repo — enabling and re-arming auto-merge, reading mergeStateStatus correctly, triaging CI failures from per-job logs, classifying transient vs real failures, resolving bot and human review threads (including the async-bot orphan trap), r

• fast-forwarding-branches

  Keep the primary checkout's integration branch current with a safe, interval-driven `git pull --ff-only` — and alert, never improvise, when the fast-forward is refused. Because agents work in worktrees cut from the primary, a stale primary silently hands every new worktree a stale base (stale hooks,

• figma-specs

  Tame design drift with the Figma MCP — snapshot a Figma file into living design specs anchored to STABLE identity (node IDs and published component/style/variable keys) rather than page/frame names or position, so when designers reorganize the file the links still resolve. Each component is mapped t

• figma

  Connect to Figma from your AI tool — read design context (layout, typography, colors), fetch variables/tokens, capture visual references, generate code from frames, and optionally map Figma components to code with Code Connect. Use whenever a task involves a Figma file, design tokens, a design-to-co

• gh-issue-filing

  Write and triage GitHub issues that AI sub-agents can implement without asking a single follow-up question — the six-section groomed-issue format (verified file:line evidence, literal acceptance criteria, traceability hint), the filing-time label taxonomy, duplicate and cross-reference etiquette bet

• gh-issue-labels

  Defines and operates the label-based control plane for multi-agent GitHub orchestration — agent-model:* (per-issue model override), agent-effort:* (effort directive carried in the dispatch prompt, not an API param), agent-claimed (active lock marker), do-not-dispatch (operator hold), do-not-rebase (

• gh-issue-locking

  Claim/lock/release protocol for GitHub issues worked by AI agents, with multi-operator coordination etiquette. Use whenever an orchestrator is about to claim, dispatch, or start work on an issue (lock BEFORE dispatch, always); whenever a PR linked to a claimed issue merges or closes (release the loc

• grooming-issues

  Interactively groom an issue that is not yet ready for development — engage the engineering/project lead with questions to assert the intent, scope, and direction of the ask until ≥90% certain the written story reflects what they want, then fill the issue body to the six-section groomed anatomy. Cap

• humanizer

  Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive 'Signs of AI writing' guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyse

• jira-issue-fields

  Defines and operates the field- and label-based control plane for multi-agent Jira orchestration — issue type (bug/story/task selection), the native Priority field (P0→Highest … P3→Lowest), agent-model-* labels (per-issue model override), agent-effort-* labels (effort directive carried in the dispat

• jira-issue-filing

  File and triage Jira issues that AI sub-agents can implement without asking a single follow-up question — the six-section groomed-issue format (verified file:line evidence, literal acceptance criteria, traceability hint) in the issue description, the filing-time issue-type/Priority/label taxonomy, d

• jira-issue-locking

  Claim/lock/release protocol for Jira issues worked by AI agents, with multi-operator coordination etiquette. Use whenever an orchestrator is about to claim, dispatch, or start work on a Jira issue (lock BEFORE dispatch, always); whenever a PR linked to a claimed issue merges or closes (release the l

• mcp-builder

  Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

• orchestrating-slots

  Run an N-slot, event-driven multi-agent orchestration loop over a repo — survey PRs and issues, fill concurrency slots with sub-agent dispatches P0-first, react to merge/CI/review events, apply backpressure when the queue jams, and keep all state in labels/fields, assignees, timestamps, and a sentin

• pdf

  Use this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, ext

• pptx

  Use this skill any time a .pptx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx file (even if the extracted content will be used elsewhere, like in an email or summary); e

• pr-checks

  Single-lane interval watcher that keeps open agent PRs green and mergeable — the CI + conflict + merge-queue medic lane of a multi-agent fleet. Watches for failing required checks, DIRTY/conflicted branches, AND merge_group (queue) failures that head checks are blind to; dispatches fixes, dequeues +

• pr-cleanup

  Single-lane interval watcher that tidies up after agent PRs close — the post-merge janitor lane of a multi-agent fleet. Acts ONLY on closed PRs — deletes merged local branches, removes spent worktrees, closes the linked GitHub/Jira issue on merge, and releases the claim lock. Critical when PRs land

• pr-comments

  Single-lane interval watcher that drives PR review threads to resolved, then arms auto-merge — the review-thread-closer lane of a multi-agent fleet. Watches every open agent PR for AI-review-bot and human review comments, dispatches a fix sub-agent or posts a reasoned counter, resolves the thread on

• skill-creator

  Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better t

• spelunking-specs

  Deep-dive an unknown or evolving codebase and distil it into living spec documents — spawn agents (one per discovered domain) to ingest the code, infer the conventions, standards, assumptions, and invariants, verify them against real file:line evidence, and write them as numerically-referenced domai

• technical-recon

  Technical recon of an already-groomed, stakeholder-accepted issue that the dev team has not yet vetted or sized — dispatch a read-only sub-agent (one per issue, in parallel for a batch) to trace the ask into the codebase, produce a verified implementation approach, assert a level-of-effort (LoE) est

• token-bleed-hunting

  The authoritative method for hunting TOKEN BLEED — tokens an AI coding agent spends that buy no work — from OpenTelemetry metrics and events (Claude Code first-class) and local session transcripts, then fixing it forward in the repo's harness config so every future session is cheaper at the same cap

• xlsx

Slash commands (31)

• /agents-skills:dispatch

  Start an event-driven orchestration session — fill N concurrency slots with sub-agent dispatches, split across dev slots (new issues), rescue slots (CI failures, review threads, stuck PRs), and flex slots (dev work that yields to incidents). Example — /dispatch 8, or /dispatch dev:5 rescue:1 flex:2

• /agents-skills:figma-init-spec

  Snapshot a Figma file into living design specs anchored to stable identity (node IDs, component/style keys) with a repo-owned design→code map, so reorganization doesn't break the links. Code Connect optional. For taming design drift. Optional scope arg. Example — /figma-init-spec, or /figma-init-spe

• /agents-skills:figma-refresh-spec

  Re-read Figma and diff it against the design specs by stable id, then report drift — separating pure reorganization (no code change) from real token/size/state changes (a precise re-touch list routed to code via Code Connect). Run after designs or the Figma file get moved around. Optional scope arg.

• /agents-skills:gh-issue-groom

  Groom a not-yet-dev-ready GitHub issue (pre-technical-recon) — drop one or more issue numbers and dispatch an agent per issue to ask the engineering/project lead about intent, scope, and direction until ≥90% certain the written story reflects the ask, then fill the body to the six-section groomed an

• /agents-skills:gh-issue-recon

  Technical recon of groomed-but-unvetted GitHub issue(s) — drop one or more issue numbers and dispatch a read-only sub-agent per issue (in parallel) to trace the ask into the codebase, produce a verified implementation approach, assert a level-of-effort estimate with confidence, surface risks/depende

• /agents-skills:gh-issue-use-aws-browser

  The full-stack deep dive — /gh-issue plus BOTH the live-browser and read-only AWS diagnosis phases, correlated: reproduce the symptom as the signed-in user, capture the failing request, then chase that exact request through the cloud (edge, logs, service, data) until the fault layer is found. Use fo

• /agents-skills:gh-issue-use-aws-chrome

  Like /gh-issue-use-aws-browser (correlated browser + read-only AWS full-stack dive), but the browser phase is pinned to the Chrome MCP (claude-in-chrome). Example — /gh-issue-use-aws-chrome \"file export > Download\" clicking Download 403s on staging effort:max

• /agents-skills:gh-issue-use-aws-cypress

  Like /gh-issue-use-aws-browser (correlated browser + read-only AWS full-stack dive), but the browser phase is pinned to the Cypress MCP. Example — /gh-issue-use-aws-cypress \"file export > Download\" clicking Download 403s on staging effort:max

• /agents-skills:gh-issue-use-aws-playwright

  Like /gh-issue-use-aws-browser (correlated browser + read-only AWS full-stack dive), but the browser phase is pinned to the Playwright MCP. Example — /gh-issue-use-aws-playwright \"file export > Download\" clicking Download 403s on staging effort:max

• /agents-skills:gh-issue-use-aws

  Same as /gh-issue, plus an AWS deep-dive — diagnose the reported symptom through the operator's live AWS CLI session using strictly read-only calls (logs, metrics, resource state, CloudTrail) before filing the groomed GitHub issue with cloud-traced evidence. Example — /gh-issue-use-aws "file export"

• /agents-skills:gh-issue-use-browser

  Same as /gh-issue, plus a live-browser deep-dive using whatever browser MCP is connected (Chrome, Playwright, or Cypress): reproduce as the signed-in user, watch network + console, capture the failing request/response, mock to isolate, and record a repro before filing the GitHub issue. Example — /gh

• /agents-skills:gh-issue-use-chrome

  Like /gh-issue-use-browser, but pinned to the Chrome MCP (claude-in-chrome) — use when you want your real signed-in Chrome session for the repro. Example — /gh-issue-use-chrome \"Composer > Save\" saving toasts failure but the record persists

• /agents-skills:gh-issue-use-cypress

  Like /gh-issue-use-browser, but pinned to the Cypress MCP — use when your project runs the Cypress MCP for browser automation. Example — /gh-issue-use-cypress \"Composer > Save\" saving toasts failure but the record persists

• /agents-skills:gh-issue-use-playwright

  Like /gh-issue-use-browser, but pinned to the Playwright MCP — use for a clean automated browser (public page or test credentials, no real session needed). Example — /gh-issue-use-playwright \"Composer > Save\" saving toasts failure but the record persists

• /agents-skills:gh-issue

  File a groomed, dispatchable GitHub issue from a breadcrumb (component/page/object) plus a problem description — recon the code, verify evidence, dedupe, label, file. Optionally pin the dispatch model/effort. Example — /gh-issue "Settings > Billing > Invoice row" clicking an invoice opens a 404 mode

• /agents-skills:jira-issue-groom

  Groom a not-yet-dev-ready Jira issue (pre-technical-recon) — drop one or more issue keys and dispatch an agent per issue to ask the engineering/project lead about intent, scope, and direction until ≥90% certain the written story reflects the ask, then fill the description to the six-section groomed

• /agents-skills:jira-issue-recon

  Technical recon of groomed-but-unvetted Jira issue(s) — drop one or more issue keys and dispatch a read-only sub-agent per issue (in parallel) to trace the ask into the codebase, produce a verified implementation approach, assert a level-of-effort estimate with confidence, surface risks/dependencies

• /agents-skills:jira-issue-use-aws-browser

  The full-stack deep dive — /jira-issue plus BOTH the live-browser and read-only AWS diagnosis phases, correlated: reproduce the symptom as the signed-in user, capture the failing request, then chase that exact request through the cloud (edge, logs, service, data) until the fault layer is found. Use

• /agents-skills:jira-issue-use-aws-chrome

  Like /jira-issue-use-aws-browser (correlated browser + read-only AWS full-stack dive), but the browser phase is pinned to the Chrome MCP (claude-in-chrome). Example — /jira-issue-use-aws-chrome \"file export > Download\" clicking Download 403s on staging effort:max

• /agents-skills:jira-issue-use-aws-cypress

  Like /jira-issue-use-aws-browser (correlated browser + read-only AWS full-stack dive), but the browser phase is pinned to the Cypress MCP. Example — /jira-issue-use-aws-cypress \"file export > Download\" clicking Download 403s on staging effort:max

• /agents-skills:jira-issue-use-aws-playwright

  Like /jira-issue-use-aws-browser (correlated browser + read-only AWS full-stack dive), but the browser phase is pinned to the Playwright MCP. Example — /jira-issue-use-aws-playwright \"file export > Download\" clicking Download 403s on staging effort:max

• /agents-skills:jira-issue-use-aws

  Same as /jira-issue, plus an AWS deep-dive — diagnose the reported symptom through the operator's live AWS CLI session using strictly read-only calls (logs, metrics, resource state, CloudTrail) before filing the groomed Jira issue with cloud-traced evidence. Example — /jira-issue-use-aws "file expor

• /agents-skills:jira-issue-use-browser

  Same as /jira-issue, plus a live-browser deep-dive using whatever browser MCP is connected (Chrome, Playwright, or Cypress): reproduce as the signed-in user, watch network + console, capture the failing request/response, mock to isolate, and record a repro before filing the Jira issue. Example — /ji

• /agents-skills:jira-issue-use-chrome

  Like /jira-issue-use-browser, but pinned to the Chrome MCP (claude-in-chrome) — use when you want your real signed-in Chrome session for the repro. Example — /jira-issue-use-chrome \"Composer > Save\" saving toasts failure but the record persists

• /agents-skills:jira-issue-use-cypress

  Like /jira-issue-use-browser, but pinned to the Cypress MCP — use when your project runs the Cypress MCP for browser automation. Example — /jira-issue-use-cypress \"Composer > Save\" saving toasts failure but the record persists

• /agents-skills:jira-issue-use-playwright

  Like /jira-issue-use-browser, but pinned to the Playwright MCP — use for a clean automated browser (public page or test credentials, no real session needed). Example — /jira-issue-use-playwright \"Composer > Save\" saving toasts failure but the record persists

• /agents-skills:jira-issue

  File a groomed, dispatchable Jira issue from a breadcrumb (component/page/object) plus a problem description — recon the code, verify evidence, dedupe via JQL, set fields, file via the Atlassian MCP. Optionally pin the dispatch model/effort. Example — /jira-issue "Settings > Billing > Invoice row" c

• /agents-skills:kill

  Emergency load-shedding when sub-agents overload the machine — OOM, runaway pre-push hooks, memory-compressor or swap pressure. Kills the newest agents first (least invested work) until pressure stabilizes, then salvages their worktrees. Example — /kill, /kill 3, /kill all

• /agents-skills:rescue

  Rescue a stuck PR or a failing environment — diagnose and fix failing CI/CD, PR checks, blocking review threads, and merge conflicts, then re-arm the merge path. Example — /rescue 1234, or /rescue dev

• /agents-skills:spelunking-init-spec

  Deep-dive the codebase and distil it into living, numerically-referenced domain specs — spawn survey agents to auto-discover the domains, fan out one agent per domain to infer conventions/standards/assumptions/invariants with verified file:line evidence, verify the claims, and write the spec set (sp

• /agents-skills:spelunking-refresh-spec

  Refresh the living codebase specs after the code has moved — re-recon a spec reference, a domain, a path, or (no arg) the stalest specs; diff each against current code, update discovered behaviour and conventions in place, transition confidence with evidence, flag regressions as drift, append new re

Subagents (18)

• branch-ff

  Fast-forward lane. Loops on an interval and keeps the PRIMARY checkout's integration branch at the remote tip with a safe `git pull --ff-only` — so every worktree an agent cuts starts from a current base (hooks, scripts, skills). Purely mechanical — run the tick, match the decision table, report one

• builder

  Surgical 1–2 file edit — typo fixes, single-function changes, mechanical renames, format-preserving tweaks. Hard-refuses 3+ file scope, new files (unless asked), or new features. Returns a tight diff receipt. Use for bounded, obvious edits; not for cross-file refactors or anything needing design.

• ci-flake-hunter

  CI-stability medic and the SOLE flake hunter for a repo. Continuously mines CI history for real flakes (shards that intermittently red the per-PR gate or the merge_group speculative build and eject correct PRs), reproduces them harsher-than-CI, root-causes the real race, and fixes them FORWARD — nev

• ci-speed-hunter

  CI-latency medic and the SOLE speed hunter for a repo. Continuously mines recent CI history (per-PR gate + merge_group) to find where CI is SLOW — long-pole jobs, the wall-clock critical path, under-sharded matrices, cold caches, and tier gates that fail to skip work — then makes CI faster WITHOUT l

• deploy-watcher

  Deploy-health medic watcher lane. Loops on an interval over the integration environment's deploy workflow (deploy step, migrations/bootstrap, post-deploy smoke) and keeps it green by fixing FORWARD — never reverts, never skips tests, never masks. Triages transient (concurrent-deploy state-lock race,

• design-mapper

  Per-domain Figma design mapper for /figma-init-spec and /figma-refresh-spec — pulls a component set or screen flow via the Figma MCP (get_variable_defs, get_code, get_image) and writes/updates its design specs anchored to stable node IDs and component keys, with a repo-owned design→code map found by

• diagnostician

  Read-only repro and root-cause worker for the /…-use-aws and /…-use-browser commands and recon — drives the browser via whichever browser MCP is connected (Chrome, Playwright, Cypress, or Chrome DevTools) in its own context and/or digs AWS logs, config, and CloudTrail with strictly read-only calls,

• implementer

  Implements one groomed issue end-to-end — works in its own worktree/branch, writes tests first, keeps scope to the issue, runs the pre-push gate, opens a PR with auto-merge armed, and returns the completion contract. The dev agent /dispatch spawns into a dev slot.

• pr-checks

  CI + conflict + merge-queue medic watcher lane. Loops on an interval over open agent PRs, fixes failing required checks and DIRTY/conflicted branches, and runs full merge-queue health every tick — head-green is not queue-green. Dequeues + drafts poison-pill PRs that drop an ALLGREEN group, and re-ar

• pr-cleanup

  Post-merge janitor watcher lane. Loops on an interval over CLOSED agent PRs only — deletes merged local branches, removes spent worktrees, closes the linked GitHub/Jira issue on merge, and releases the claim lock. A PR closed without merging frees the lock but leaves the issue open and re-dispatchab

• pr-comments

  Review-thread-closer watcher lane. Loops on an interval over open agent PRs, drives bot + human review threads to resolved (fix or reasoned counter, resolve only after the fix is pushed), and owns the initial arm-after-bot-review of auto-merge. Re-arming after a merge-queue ejection is the pr-checks

• pr-dispatcher

  Slot-filler watcher lane — the continuous form of `/dispatch`'s dev slots. Each tick it claims groomed, unblocked, unclaimed issues, spawns an `implementer` per issue, tracks each to an open PR, hands that PR to the watcher lanes, and refills the freed slot. Honors the label control plane (do-not-di

• pr-rescuer

  Unsticks a stuck PR — reads mergeStateStatus, triages CI from per-job logs (transient vs real), resolves bot and human review threads, recovers DIRTY/conflicted branches, and re-arms auto-merge. The worker /rescue and dispatch rescue slots use. Diagnoses before it mutates.

• reviewer

  Adversarial diff/branch/PR reviewer. One line per finding, severity-tagged, no praise, no scope creep, format `path:line: <severity>: <problem>. <fix>.`. Use for 'review this PR/diff/file'. Skips style nits unless they change meaning.

• scout

  Read-only code locator. Given "where is X", "what calls Y", "list uses of Z", or "map this area", returns a tight file:line map in caveman shorthand and nothing else — no fixes, no opinions. Use to offload mechanical lookups off the main session onto the cheap tier (see MODEL-DEFAULTS.md).

• spelunker

  Per-domain codebase mapper for /spelunking-init-spec and /spelunking-refresh-spec — runs the spelunk loop (framework introspection + machine-name following + runtime observation) over one assigned domain and writes its numbered spec files with verified file:line / command evidence. Writes only under

• token-bleed-hunter

  Token-efficiency medic and the SOLE token-bleed hunter for a repo. Continuously mines OpenTelemetry metrics and events from AI coding-agent sessions (Claude Code first-class), plus local session transcripts, to find token BLEED — spend that buys no work — then fixes it forward in the repo's harness

• verifier

  Verify-then-trust gate. Confirms a claimed result actually holds before it's trusted — the PR exists and is the linked one, required CI is green, acceptance criteria are met, and any cited file:line / recon / spec claim is real. Returns a pass/fail verdict with evidence. Read-only; an unverifiable c

Similar plugins