shared

Deep, iterative, evidence-based analysis of situations, decisions, trade-offs, and open questions. Use when asked to analyze, evaluate, think through, assess implications, weigh options, or explore a problem space in depth before acting. Triggers: analyze, think through, evaluate, implications, trade-offs, help me think, considerations, second-order effects, what would happen if, assess, examine, pros and cons, decision analysis, situation analysis.

Evidence-based evaluation of AI-generated findings — review feedback, audit findings, spec challenges, or any AI-generated assessment that requires triage before action. Teaches agents to investigate before accepting, calibrate confidence based on evidence quality, classify findings into actionable categories, and detect assessment bias. Load this skill when evaluating findings from reviewers, auditors, challengers, or any AI-generated feedback loop.

Verify artifact quality by running coherence lenses (logical consistency) and factual verification tracks (external accuracy). Designed to be loaded by a freshly spawned subagent or nested Claude instance that reads the artifact cold. Produces a findings file — does not fix issues. Use when: after completing a research report, after spec scope freeze, on-demand quality check, verify claims, fact-check, review accuracy, cross-check assertions, post-completion verification for any artifact with evidence-backed claims.

Inkeep brand identity — principles, logo rules, typography, color usage, text style rules, size minimums. Background knowledge for visual creation skills (graphics, motion-video, gslides). Loaded automatically by consumer skills.

Use when: strategic tension, portfolio decisions (create/kill/pause bets), product identity, competitive response, resource allocation, direction-setting. Externalizes and sharpens product strategy through forcing questions, autonomous investigation, and deliberate disruption of fixed thinking.

Write and execute scripts that call MCP tools programmatically via @modelcontextprotocol/sdk. Use when orchestrating multiple MCP tool calls (3+), batching operations, or chaining results across tools/servers. Triggers on: multi-tool workflow, batch MCP calls, code mode, programmatic MCP, chain tool results, cross-server orchestration.

Consolidate factual information from multiple source documents into one distilled body of work while maintaining factual fidelity. Source-agnostic: works on agent outputs, articles, books, web search results, structured or unstructured. Optionally scoped to a criteria/goal/purpose. Surfaces conflicts, ambiguities, and open questions transparently. Use when merging fan-out agent outputs, combining research sub-reports, consolidating articles/books, or any multi-source knowledge fusion task. Triggers on consolidate, merge reports, combine findings, fan-out convergence, multi-source synthesis.

Search, navigate, and resume Claude Code conversation histories. Find conversations by topic, PR, branch, worktree, skill used, files modified, repo, or recency. Use when asked to find a prior conversation, recall past work, resume interrupted sessions, or search chat history. Triggers on find conversation, prior conversation, previous session, resume session, where did we, what was I working on, find the chat, search history.

Upload media files (video, images, GIFs) to Bunny CDN or Vimeo. Use when a skill needs to publish rendered videos, screenshots, or other media to a permanent URL. Provides three upload functions for different platforms and use cases. Triggers on "upload video", "publish to vimeo", "upload to bunny", "publish media", or when another skill needs media hosting.

Shape a strategic bet and decompose it into spec-able stories — multi-multi-dimensional value per story, dependency graphs, cross-cutting concerns, phasing with rationale. Use when: decomposing a bet into stories, structuring a product direction into workable pieces, mapping dependencies across workstreams, preparing stories for /stories or /spec.

Conduct technical research and produce formal reports by default. Can also deliver findings directly or update/extend an existing report. Reports default to <repo-root>/reports/ when in a git repo, or ~/reports/ otherwise. Use when investigating technologies, comparing systems, analyzing codebases, documenting architectures, gathering context for decisions, or when asked to refresh/update prior research.

Validate whether a SaaS platform's APIs can be accessed using browser session cookies — testing feasibility of a Chrome extension session proxy. Connects to user's authenticated Chrome, enumerates cookies, captures network traffic, tests official and internal APIs with cookie auth, checks Origin/CSRF requirements, and produces a structured findings document with a credential extraction recipe. One platform at a time. Triggers: session recon, cookie auth testing, extension proxy feasibility, SaaS API cookie test, session proxy validation.

Capture, annotate, and include screenshots in pull requests for UI changes. Use when creating or updating PRs that touch frontend components, pages, or any web-facing surface. Also use when asked to add before/after screenshots, visual diffs, or enrich PR descriptions. Triggers on: PR screenshots, before/after, visual diff, PR description, capture screenshot, PR images, enrich PR.

Sharpen a single work item into a /spec-ready story seed — problem framing (SCR), multi-dimensional value with intersection reasoning, falsifiable invariants, temporally-tagged non-goals, acceptance criteria, assumptions with verification plans. Use when: refining a story before /spec, sharpening an ad-hoc idea into a structured seed, enumerating invariants and non-goals for a work item, preparing a handoff from product thinking to technical specification.

Shared vocabulary and protocols for structured decision-making: decision taxonomy (LOCKED/DIRECTED/DELEGATED, NEVER/NOT NOW/NOT UNLESS, confidence labels), artifact discipline (progressive writing, evidence conventions, dual-audience design), problem framing (SCR format, stress-test probes), disambiguation protocol (challenge/probe/surface/verify), challenge posture (co-driver stance, anti-sycophancy, investigate-vs-judgment boundary), value dimensions (multi-dimensional articulation, intersection reasoning, dimension-trace diagnostic). Load when doing product planning, strategy, specs, or producing durable artifacts.

Multi-channel grounding and landscape discovery. Mechanically harvests all available knowledge channels (web probes, existing reports, codebase exploration, catalog skills, OSS repos, user-provided sources) and synthesizes findings into one structured world model. Non-prescriptive — reports observations, convergences, divergences, and gaps without recommending. Use before spec work, research, analysis, or any task that needs comprehensive situational understanding. Triggers: worldmodel, world model, ground this, map the landscape, build context, what do we know about, understand this topic, grounding, context gathering.

Design and write high-quality Claude Code agents and agent prompts. Use when creating or updating .claude/agents/*.md for (1) single-purpose subagents (reviewers, implementers, researchers) and (2) workflow orchestrators (multi-phase coordinators like pr-review, feature-development, bug-fix). Covers delegation triggers, tool/permission/model choices, Task-tool orchestration, phase handoffs, aggregation, iteration gates, and output contracts. Also use when deciding between subagents vs skills vs always-on repo guidance.

Create or revise Claude Code-compatible Agent Skills (SKILL.md with optional references/, scripts/, and assets/). Use when designing a new skill, improving an existing skill, or updating/refactoring an existing skill while preserving the original author's intent (avoid semantic drift unless explicitly requested/approved by the author). Also use when integrating skills with subagents (context fork, agent).

1 hook across 1 event