Agent
Read-only second opinion at a different model tier, for architecture-scoped plans, stuck-debugging escalation, and ad-hoc algorithmic/logic decisions. Invoked explicitly per rules/architecture-review.md — not matched via delegation keywords.
Skill Claude CodeCodex
Exhaustive audit of your Claude Code configuration against the official docs of the latest version: settings, permissions, hooks, agents, rules, skills, mcp, statusline, output-styles, memory. Detects non-conformities, staleness, over-engineering, cross-file inconsistencies; produces a correct / missing / incorrect…
Skill Claude CodeCodex
Analyzes the Claude Code release notes since the last processed version and produces, in tmp/, a ready-to-execute plan to adapt the project config (.claude/settings.json, hooks, permissions, rules, agents, skills, CLAUDE.md) to the relevant novelties. Bootstraps the last N releases on the first run. Keywords…
Skill Claude CodeCodex
Pure functional review: verify that the implemented business logic is the right one among all possible logics — grounded in Fagan Inspections, Decision Table Testing, and Mutation Analysis thinking. Keywords: functional review, business logic, edge cases, decision table, Fagan inspection, mutation analysis, is this…
Skill Claude CodeCodex
Issue management with automatic project attachment and epic/sub-issue linking. Keywords: issue, epic, sub-issue, ticket, backlog, project board, gh issue create. Apply when creating tracked issues or organizing epics.
Skill Claude CodeCodex
Analyze a meeting/call transcript and produce a literature-grounded, actionable summary committed to the repo. Keywords: meeting summary, transcript, minutes, action items, decisions, debrief, call notes, standup, BLUF, actionable summary, follow-up. Apply when the user provides a transcript (pasted text or file) of a…
Skill Claude CodeCodex
Multi-agent local code review (pre-commit / pre-PR) that fans out independent reading perspectives over a diff, anchors every finding to verbatim evidence, and escalates to adversarial verification on high-risk changes. Distinct from a platform's built-in PR review: this reviews uncommitted or branch-local changes…
Skill Claude CodeCodex
Start a tracked issue with discipline: fetch it, run a Definition-of-Ready gate, detect likely duplicates, check its place in the epic/sub-issue hierarchy, decide branch vs. isolated worktree, mark it in-progress, write a short briefing, then hand off to planning. Keywords: start ticket, handle ticket, work on issue…
Skill Claude CodeCodex
Sync your project's docs from a GIT REFERENCE (commit-range driven): analyzes the commits since a ref, proposes an update/delete/create plan in a scratch file, waits for validation, then applies it. Covers docs/, CLAUDE.md (or equivalent), .claude/rules/, .claude/skills/, and README files across the repo. Keywords…