trilogy-group

Use when the user asks to respond to Greptile PR review comments, address Greptile feedback, or handle automated code review comments on a pull request

Use to build new features or fix capabilities. Orchestrates design, build, and verification across three layers.

design

Use after the external spec of the domain is know and before writing domain code. Internal design process for domain-builders. Guides file structure, design guard creation, and internal decomposition decisions before writing implementation code.

Single authority for database migration creation, ordering, and conflict prevention

System Documentation Generator (Collaborative)

Use repo-local GraphQL helpers to read and write Linear through `LINEAR_API_KEY`.

Use this skill when a repository contains a markdown implementation plan that needs to become real Linear issues with correct parent/sub-issue relationships, blockers, and Definition of Ready context.

Generate a structured implementation plan from project requirements with decomposed tasks, milestones, dependencies, and acceptance criteria. Use when starting a new project to create docs/tasks/*.md files suitable for conversion to Linear issues.

Land a PR by monitoring conflicts, resolving them, waiting for checks, and squash-merging when green; use when asked to land, merge, or shepherd a PR to completion.

Bootstrap, validate, and scale manifest-driven product intelligence systems that combine Obsidian, source ingestion, code intelligence, and recurring automations. Use when a user wants to onboard one product or many products into a repeatable second-brain workflow with low incremental human effort.

Scaffold, audit, and evolve an Obsidian vault into a reusable intelligence system or product second brain. Use when a user wants to turn project documents, repos, links, notes, PDFs, or recurring entity records into a linked vault with YAML frontmatter, internal links, dashboards, templates, Bases, Canvas, hub nodes, intelligence summaries, and research-to-action synthesis.

Build and audit engineering-ready product intelligence around services, local runtime, CI/CD, deployments, support-to-code workflows, and scale-readiness. Use when a user wants a product second brain that engineers can actually ship, debug, deploy, and support from.

Harden a fresh Ubuntu 24.04 box for production OpenClaw use. UFW rate-limit, sshd hardening (modern KexAlgorithms/Ciphers/MACs, AllowUsers, MaxAuthTries), fail2ban with a local jail, unattended-upgrades with auto-reboot, sysctl kernel tuning, optional TOTP MFA on SSH, systemd lingering for the target user, plus a structured credential blast-radius conversation (dedicated bot Gmail, dedicated social handles, spend caps, never-on-this-box list). Tracked in beads. Skill 2 of 3. Requires /openclaw-prereqs to have run first.

Install OpenClaw on a hardened Ubuntu 24.04 box. Installs Claude Code + cc-openclaw skills, OpenClaw gateway as a systemd user service, secrets via ${ENV_VAR} placeholders, Telegram bot (with privacy-mode gotcha), Xvfb + Chrome headless browser stack, runtime beads at .openclaw/workspace/.beads, optional GNU Stow dotfiles, SSH tunnel instructions, first cron job, and a final audit. Tracked in beads. Skill 3 of 3. Requires /openclaw-harden to have run first.

Install bootstrap prerequisites (Node 24, beads, jq) on a fresh Ubuntu 24.04 box and initialize the shared beads task graph that openclaw-harden and openclaw-install will track their work in. Idempotent. Run this first on any new box. Skill 1 of 3.

The complete, self-contained brief and operator manual for the live Grade-5 reading course EduLLM Grade 5 Reading (sourcedId g5-reading-merged-pp-9802; PowerPath-100, alpha.timeback.com). This folder is fully portable — SKILL.md plus co-located references/ (per-item ledger, evidence summary, decision memo, BrainLift, dated pull record, cover email) and scripts/ (ten stdlib-only Python tools including a self-contained TimeBack API client); no repository access required. It serves two audiences from one file. (1) Evaluators and reviewers — load this when asked to "review g5-reading-merged-pp-9802", "evaluate g5-reading-merged-pp-9802", "judge the grade 5 reading course", "access the G5 course", "open the course", or assess the G5 reading bake-off entry; it carries the access path (sign-in, the enrolled reviewer accounts, the blank-course fallback), the full design argument (the leak-by-format finding, per-item cross-family certification, the conversion-by-rewrite result), the live-verified build state (94 tests

The knowledge base a REVIEWER AGENT (e.g. an academics-team evaluating agent) loads to understand and evaluate the Incept MAP Proxy — an adaptive G2-G5 reading CAT designed to place students on the same ability continuum as NWEA MAP Growth Reading 2-5. Covers the instrument's defining feature of mastery, its BrainLift (the seven load-bearing Spiky POVs and their evidence), what is actually built and live-verified (the authoritative v4 full-bank simulation suite, 2026-07-03, on the full 786-item bank), the case for approval (a Stage-2 feasibility pilot), the audit history (two adversarial audits + a teammate review, and every finding's disposition), seven cross-source syntheses that triangulate the BrainLift + Case for Approval + Audits together, the honest tradeoff logic (locked decisions vs. genuinely open forks vs. deliberately dropped approaches), and the full limitations/honest-ceiling disclosure. Use when a reviewer agent is asked to "review/evaluate the MAP proxy", "decide whether to approve the MAP pro

Guided, open-ended self-assessment that captures HOW a person uses AI — the tools/clients they work in (Claude Code, Codex, OpenCode, Cursor, ChatGPT/Claude web, custom agents/scripts), the providers and models they call, and the ACCESS PATH (direct API key, TrueFoundry gateway/VAT, a Max/ChatGPT subscription, or cloud like Bedrock/Vertex). Emits a STRUCTURED, schema-first report the AI CoE can aggregate to find cost-optimization opportunities. Use when someone says "assess my AI usage", "AI cost assessment", "the AI CoE asked me to document how I use AI", or is asked to describe their AI usage for cost review. Does NOT assume everyone organizes work into "workflows" — it captures usage by tool + access path, and only drills into production pipelines when they exist.

Canonical template for authoring a Trilogy AI CoE skill. Use as the starting point when creating a new skill for the aicoe-skills marketplace — copy this folder, rename it, rewrite the frontmatter description, and replace the body with your skill's instructions.

Advanced code review skill with multi-language support and best practices enforcement

PDF processing skill for extraction, analysis, and manipulation of PDF documents

Interactively discover patentable ideas in a codebase and generate high-quality patent disclosure documents. Trigger when the user says "patent disclosure", "invention disclosure", "what's patentable", "find novel ideas", "generate disclosure", or invokes /patent-disclosure.

Deep research agent that decomposes questions into a dependency graph, researches in parallel, verifies findings, and produces cited reports.

Package QTI items + a course skeleton into a complete TimeBack Alpha Read course (OneRoster v1p2 + QTI 3.0) that passes the offline contract validator, or one-command-build a whole course from a skeleton. Use when assembling, validating, or preparing Grade-3 Alpha Read content for upload, or ingesting an incept-qti-sdk output_dir. Reverse-engineered from a live production Alpha Read course export (1077 items) and cross-checked against Ilma's timeback skill.

Submit, list, monitor, and cancel Fireworks fine-tuning jobs (SFT, DPO, RFT) for the Trilogy team via the fair-scheduler queue. ALWAYS use this instead of calling the Fireworks fine-tuning API directly or running `firectl supervised-fine-tuning-job ...`, `firectl dpoj ...`, or `firectl rft-job ...`.

Install and maintain the Burst dashboard Cursor hook (send-to-dashboard telemetry). Use when the user mentions Burst, burst.trilogy.com, AI coding activity dashboard, or their Cursor hook was lost after wseng-skills install.