This repository is a collection of Agent Skills that guide AI coding agents through repeatable engineering workflows. Each skill is a self-contained SKILL.md file under skills/<skill-name>/ with YAML frontmatter (name, description, and optionally allowed-tools) followed by Markdown instructions the agent loads on demand.

The skills are aimed at agent and developer-tooling workflows, and several are built around Sourcegraph Code Search, Deep Search, and Batch Changes:

• update-repo-docs — Aligns or creates a root README.md, summarizes related repositories and dependencies, and creates CODEOWNERS only when absent.
• create-batch-change — Drafts, scopes, and validates Sourcegraph Batch Changes batch spec YAML using Code Search queries and Sourcegraph MCP tools.
• architecture-alignment — Evaluates repositories against FINOS architecture-as-code and CALM patterns using local artifacts rather than CALM Hub.
• research-and-understanding — Uses Sourcegraph Deep Search for task scoping, implementation planning, and review across one or many repositories.

No build, package manager, or runtime dependencies are required. The repository contains only Markdown skill definitions.

git clone github.com/trly/skills
cd skills

Each skill lives in its own directory and is defined by a single SKILL.md:

skills/
  update-repo-docs/SKILL.md
  create-batch-change/SKILL.md
  architecture-alignment/SKILL.md
  research-and-understanding/SKILL.md

Skills are consumed by an Agent Skills-compatible host: the agent reads the frontmatter name and description to decide when a skill applies, then loads the skill body for detailed instructions. Skills that declare allowed-tools require those tools to be available in the agent environment (see System context).

These external dependencies are referenced directly by individual skills and are needed to fully use them: