Skill Claude Code
Write copy in a specific human voice using a three-role pipeline. Use when the user wants copy that "sounds like me", "sounds human", "matches our brand voice", "doesn't read like AI", "rewrite this in my voice", "humanize this draft", "make this less ChatGPT-ish", hands over writing samples for new copy in that…
Skill Codex
Interrupt a stalled AI work session and restore movement toward the original outcome. Use proactively when the agent is passive, trapped in narration, halting on tool exits, or reporting activity without milestone change, as well as when overengineering, inventing machinery around work, or reopening settled decisions.…
Skill Claude Code
The 5-and-5 decision checksum for an in-scope plan, spec, PR description, or fix. Use whenever the user asks for a decision checksum — "five", "/five", "give me the five", "key highlights", "top five", "what does this plan do", "what did you decide", "what does this NOT do", "what am I not getting" — during planning…
Skill Claude CodeCodex
Audit, harden, and optimize CI/CD pipelines across diverse repositories using production-tested, zero-cost principles. Use when an operator asks to "audit our CI/CD", "speed up test runs", "fix flaky CI", "harden pre-push gates", "design CI tiering", "eliminate test contention", or evaluate CI architecture before…
Skill Codex
Catch a repo's SOP (standard operating procedure) and lessons-learned docs up to what actually happened recently. Deep-recons the codebase for existing SOP / runbook / lessons / postmortem / incident docs (any .md, .txt, .rst), mines recent events from every available source (git, PRs, issues, incident files…
Skill Codex
Find the class of bugs that keeps coming back and file one umbrella issue to fix the root cause instead of the symptoms. Scans the last 14 days of GitHub issues, fix/revert commits, and PR review threads, clusters related bugs, ranks clusters by a composite churn score (reopens, repeat fixes on the same files, comment…
Skill Claude CodeCodex
List every skill in this repo with its name and one-line description, read live from skills//SKILL.md frontmatter. Use when the user asks "what skills are here", "list skills", "what can I run", or wants a quick catalog of the repo's skills. Read-only.
Skill Claude CodeCodex
Debug failing CI pipelines, pre-push gates, hosted GitHub Actions workflows, and reproduce test failures in safe isolation. Applies debug-mantra ground-truth discipline, recon seam/blast radius mapping, and foundationally sound (non-bandage) ponytail least-mechanism resolutions. Use when an operator says "fix failing…
Skill Claude CodeCodex
Run a seven-day governance review for an XYZ Forge or XYZ-vendored repository: check GitHub default-branch reconciliation, compare RebalanceOS HiQS and git activity with PDDA/RELEASES/XYZ metadata, file a ranked umbrella issue, correct verified metadata, and refresh the configured project kanban through its canonical…
Skill Claude CodeCodex
Re-orient to this repo's PDDA operating order. Walks the ROUTER.md startup sequence, names the canonical next file for the task at hand, and reports current PDDA state (active docs, roadmap coverage, deterministic findings). Use at inflection points — session start, task switch, after /compact, or whenever context has…
Skill Codex
Read-only, sub-agent-driven triage of the checkouts /merge-cleanup preserved (PRESERVEUNPUSHED, PRESERVEDIRTY, PRESERVEUNVERIFIED): zip each clone to a dated backup with a manifest, fan out at most three read-only agents grouped by branch family, and establish per clone — by ancestry, by per-file content against…
Skill Codex
Start marathon preparation from intake through reviewed implementation plans, honest contracts, disjoint lanes, MARATHON.yaml, preflight and dry-run. Use for "start marathon", "marathon", a marathon queue, or a viability check of marathon docs another session prepared. Route an ambiguous bare "marathon" here; ask…
Skill Claude CodeCodex
Audit registered CI test suites and recommend retention, split, nightly, quarantine, or turn-off verdicts based on runtime receipts, failure history, touch-set overlap, and prose/junk assertion analysis. Use when an operator asks to "audit test suites", "curate CI tests", "find slow tests", "identify redundant tests"…
Skill Claude CodeCodex
Keep the current agent moving on an already-authorized task while CI, tests, reviews, or other asynchronous work finishes. Arm a native same-conversation wake every 10 minutes for at most 3 hours, check live evidence, advance the nearest milestone, and cancel the wake when work ends. Use for "handsfree", "watch this…
Skill Claude CodeCodex
Assess whether a claimed problem or blocker is real, consequential, and worth fixing now. Use at the first failure that blocks a task, when investigation stops producing evidence, or when asked whether a feature, CI test, or other machinery is necessary. Ground the decision with recon, debug-mantra, and ponytail…
Skill Claude CodeCodex
Adaptive daily planning skill that pivots the active weekly plan based on what closed yesterday. Directly reuses the core engine and adversarial Northstars from weekly-planner (skills/3-weekly/weekly-planner/) to eliminate logic drift and prevent planning contradictions. Re-sequences the daily merge queue and…
Skill Claude CodeCodex
Answer "where are we at on ?" with a deep ground-truth assessment of a module, feature, or subsystem. Cuts through stale documentation and superficial LLM scans by orchestrating /recon (codebase architecture & live symbols), /debug-mantra (empirical breadcrumb observation), and /merge-cleanup (multi-checkout, active…
Skill Claude Code
Unified IDE task-list grooming for XYZ Forge: one central core with per-IDE adapters (ZCode, Antigravity, Codex desktop, Claude Code, Claude Desktop) that date-stamp task titles with the task's own last-activity mm-dd, write last-action descriptions, and manage pins under one safety contract. Use when asked to…
Skill Claude CodeCodex
Dual-horizon weekly planning and team orchestration skill built around XYZ Forge PRS (releases.db) and PDDA, with zero-dependency markdown fallback. Ingests team weekly GitHub issues, maps member focus domains, prioritizes high-value blockers, enforces the Provisional Decisions Protocol (zero stalls on external…
Skill Claude CodeCodex
Publish XYZ-forge into its standalone downstream repos (XYZ-mini, XYZ-skills-army-mini, AgentChorus-Skill) with the one embedded-manifest publisher, utils/py/xyzminisync.py (GH-589, GH-955). XYZ-forge is the upstream for all of them. Target one repo, several, or all in one run; preview first, then apply and push on…
Skill Claude CodeCodex
Load, verify or troubleshoot the read-only XYZ Claude Code mod that adds /xyz-status — recent hosted runs on development, tick claims and marathon/relay driver state, printed with no model turn. Use when the operator asks to "install the xyz mod", "turn on /xyz-status", "check the xyz mod", or to run its manual…
Skill Claude CodeCodex
Find documentation and execution follow-through gaps in the current session's available conversation. Trace findings, decisions and technical spikes into issue/PR records and next-step execution plans; return numbered possible orphaned tasks with evidence and next actions. Use for /close-the-loop, "what did we leave…
Skill Claude CodeCodex
Meticulous local code review ladder (uncommitted changes, a branch, or files) using /recon and /debug-mantra to test each fix and feature against live ground truth, then /workhorse and /unstuck to resolve reversible issues without stopping for operator permission. Reviewing a GitHub PR? Use review-pr-xyz (validates…
Skill Claude CodeCodex
Ground-truth GitHub pull request review that first confirms the PR's claims and the request behind it are valid (the bug is real, the issue asks for this, it has not already landed), then maps blast radius with /recon and tests every fix and feature with /debug-mantra. Report-only: posts graded findings to the PR and…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: