Skill Claude CodeCodex
Open a GitHub issue in e1024kb/wise-claude reporting feedback, a bug, or a suggestion about a wise skill or workflow — the submitter's environment (OS, Claude Code version, gh version, the cwd's public git remote as org/repo) is auto-captured and the Problem / Summary / Proposal sections are drafted from the user's…
Skill Claude CodeCodex
Reorient the agent after a session fork: inherited context becomes read-only background and pre-fork work stays with the original session. Every in-flight task, plan, and promise from before the fork is dropped, remembered file state is distrusted, and this session starts fresh parallel work - optionally on a new goal…
Skill Claude CodeCodex
Deep-research ANY underspecified subject — a tracker ticket (Jira, Linear, GitHub, …), a doc link (Confluence, Notion, RFC, Google Doc), or a free-form prompt / question — across every reachable source: the subject's own comments + screenshots, linked docs, Slack discussions, Google Drive, design files, the codebase…
Skill Claude CodeCodex
Human-first writing rules for EVERY outbound message composed on the user's behalf - ticket descriptions and comments in ANY issue tracker (Jira, Linear, GitHub Issues, GitLab, Asana, Monday, YouTrack, …), PR / MR descriptions, comments, and review replies on ANY code host (GitHub, GitLab, Bitbucket, …), pages and…
Skill Claude CodeCodex
Autonomously implement a written implementation plan (a PLAN-.md produced by ticket-auto / ticket-plan) in the current git working tree — the plan's task waves are phase gates, each task in a wave is handed to a fresh-context executor agent running in parallel, and every task lands as one atomic commit with per-task…
Agent
You are a Senior Software Engineer with 20+ years of experience, fluent across every major programming language, framework, frontend, backend, and infrastructure stack. You have been handed exactly one task from an implementation plan. Implement it well, then stop.
Skill Claude CodeCodex
First-time setup wizard — walk the user through installing wise's system deps (Python 3 + pyyaml/ulid/typingextensions, Node ≥22, gh CLI + gh auth login, markitdown for file-to-markdown extraction) and cache the probe results so workflow runs skip the live check. Idempotent — re-running only prompts for gaps. Invoked…
Skill Claude CodeCodex
Mine your own Claude Code session history for recurring task patterns and, once a pattern recurs across enough distinct sessions, draft it into a reusable personal skill under /.claude/skills/ — after you approve each one. This is the wise self-improvement loop: a SessionEnd hook quietly records each finished session…
Skill Claude CodeCodex
Review your learned skills for overlap and consolidate them — merge several redundant or composable skills into one aggregated skill and retire the originals (reversibly), after your approval. The "garden" pass to /wise-insights-mine's "harvest": mine creates skills from recurring sessions, refine keeps the resulting…
Skill Claude CodeCodex
Reversibly clean up the self-improvement loop, and roll it back. Snapshots then removes the auto-created (wise-managed) skills and/or the insights index (ledger, candidates, decisions) into a timestamped restore point — and can restore any restore point. Never touches hand-written skills, and never hard-deletes…
Skill Claude CodeCodex
File-to-markdown text extraction via Microsoft's markitdown CLI — the ONE tool to reach for whenever text/content must be extracted from a binary or structured file: PDF, Word (.docx), PowerPoint (.pptx), Excel (.xlsx/.xls), images (EXIF metadata), audio (metadata + transcription), HTML, CSV/JSON/XML, ZIP archives…
Skill Claude CodeCodex
Attach Copilot code review to the PR for the current branch, then prompt the user for individual reviewers (free-text comma-separated logins, with CODEOWNERS-derived candidates and org members shown inline for reference). Idempotent — already-requested reviewers are detected and not re-requested. This skill runs just…
Skill Claude CodeCodex
Autonomous variant of /wise-pr-create — detect the PR state for the current branch and create or refresh the PR with NO prompts. The body is drafted from the project's PR template; the base branch is chosen autonomously (the repo's default branch) instead of asked. Built for unattended ticket→PR runs. Invoked as…
Skill Claude CodeCodex
Detect the PR state for the current branch and either create a new GitHub PR or refresh an existing one — body drafted from the project's .github/pullrequesttemplate.md (or a bundled fallback) filled from the branch's diff + commits, Jira key auto-detected from branch/diff/log/session, base branch picked interactively…
Skill Claude CodeCodex
Autonomous variant of /wise-pr-add-reviewers — request the bot reviews on the current branch's PR with NO prompts: attach Copilot code review and trigger CodeRabbit (@coderabbitai review). It does not enumerate or ask for human reviewers. Idempotent and best-effort — a request failure never blocks. Built for…
Skill Claude CodeCodex
Autonomous variant of /wise-pr-watch — watch the current branch's PR pipelines, auto-fix failing checks (lint / tests / other), commit + push, then trigger + wait for the bot reviews (Copilot and CodeRabbit) and handle every bot review comment by severity (minors fixed, majors via a considered decision, false…
Skill Claude CodeCodex
Watch the current branch's PR and drive it to green: block on gh pr checks --watch, classify failing checks (lint / tests / other) and auto-fix them — committing via the shared Conventional-Commits routine — then walk four sequential review queues (humans → Copilot → CodeRabbit → Sonar), each gated by an interactive…
Skill Claude CodeCodex
Creates and iterates on Product Requirements Documents (PRDs) through structured discovery, parallel brainstorming agents, and collaborative refinement. Use this skill whenever the user mentions PRD, product requirements, product spec, feature spec, product brief, requirements document, or wants to define, scope, or…
Agent
You are a competitive intelligence analyst. Your job is to research the competitive landscape for a product and produce a structured report that will inform a PRD.
Agent
You are a market research analyst. Your job is to build a clear picture of the market context, user personas, and demand signals that will ground a PRD in reality rather than assumptions.
Agent
You are a senior product leader reviewing a PRD for quality, completeness, and actionability. Your review should be constructive and specific — not a rubber stamp, not a teardown, but a rigorous editorial pass.
Agent
You are a senior technical advisor. Your job is to assess technical feasibility, identify risks, and surface architecture-relevant context that should inform the PRD — without prescribing the solution.
Skill Claude CodeCodex
Set the session's token-budget profile — low, medium, or max — which every profile-sensitive wise skill and workflow reads at start to scale model tiers, optional research scope, reviewer-panel size, and retry caps. Budget only: it NEVER changes correctness rules (commit conventions, dirty-tree refusals, review gates…
Skill Claude CodeCodex
Produce a verified status report of the current session - what is done, what is not, the plan, next steps, open questions, leftovers, and postponed tasks - in compact ref-coded form (F facts, A actions, Q questions, R risks, L leftovers, P postponed), with every claim backed by checkable evidence (git SHA, PR number…