Claude Code plugins for enhanced development workflows. Includes Research → Plan → Implement workflow inspired by HumanLayer's context engineering patterns.
Agent
Extracts high-value insights from a specific document under .rpi/ — decisions made, constraints identified, technical specifications, and whether any of it is still current. Use on the few documents artifact-locator surfaces as most relevant, not as a general search.
Agent
Finds workflow artifacts in the repo's .rpi/ directory — tickets, research, designs, plans, PR descriptions — and returns them grouped by type with one-line descriptions. Use to surface prior work on a topic before researching it fresh; pair with artifact-analyzer to read the most relevant ones deeply.
Agent
Infers the ticket/issue a developer is working on from the current git branch and worktree, then fetches its details. Used by /research-codebase (and optionally /design-doc, /create-plan) as a fallback when the user runs the command without naming a ticket or question.
Agent
Analyzes HOW code works — traces data flow and explains implementation with precise file:line references. Documents the system as it exists today; does not critique, recommend, or perform root cause analysis. The more detailed your request prompt, the better.
Agent
Locates WHERE files, directories, and components live for a feature or task. Call it with a human-language description of what you're looking for — basically a "Super Grep/Glob/LS". Reports locations and organization; does not read implementations or evaluate structure.
Agent
Finds similar implementations, usage examples, and existing patterns that new work can be modeled after. Like codebase-locator, but returns concrete code snippets with file:line references, not just locations. Reports which patterns exist and how widely each is used; leaves the choice between them to the caller.
Agent
Generates focused research questions from a ticket or task description. Used as a preprocessing step by /research-codebase to keep research objective.
Agent
Researches external documentation, APIs, and technical resources on the web. Use when the answer lives outside the codebase — library docs, version-specific behavior, error messages, best-practice comparisons — or when the information is newer than training data. Returns findings with direct quotes and source links.