Sam Abbott

Create or update a CLAUDE.md project inventory for the current directory. Works at any level of the code directory hierarchy.

Review the past week's activity and suggest a prioritised plan for the coming week

Review and update project dependencies safely with incremental testing

Create a UK news summary from BBC and Guardian sources

Run tests iteratively fixing code or tests until all pass, then commit

Expert guidance for using Task (taskfile.dev) automation tool to discover and execute project development commands

Expert guidance for Stan probabilistic programming language development, including modern syntax, cmdstanr/cmdstanpy integration, and testing patterns

Open a tmux split running diffview.nvim so the user can review the agent's recent changes. Use proactively after non-trivial edits, or when the user asks to "show the diff", "open a diff viewer", or similar.

Generate or regenerate all helper scripts used by org automation skills

Scan repositories and identify issues that Claude Code can efficiently handle

Perform a code review with linting, standards checking, and priority-ranked findings

Detect active repos by the owner not cloned locally and request confirmation before acting

Analyse a GitHub repository combining activity metrics with code development insights

Analyse GitHub repository activity including issues, PRs, contributors, and engagement

Expert guidance for R package development following best practices for devtools, testthat, roxygen2, and R ecosystem tools

Expert guidance for setting up and organizing research software projects following established patterns for R packages, Julia packages, and research projects

Search for new preprints in infectious disease modelling from arXiv, medRxiv, and bioRxiv

Run the full pre-PR verification routine (format, lint, tests, type/package checks, coverage) in an isolated subagent so the main thread stays clean. Invoke this after a gh pr create is blocked by the pre-tool-use hook; report ready or unready back to the main thread.

Run as a background subagent after a PR is created or pushed. Acts as a continuous quality gate, watching CI and mergeability, and catching, answering and resolving review feedback from the review bot and from seabbs. Invoke via the Agent tool with run_in_background true.

Create a pull request based on a GitHub issue with analysis, implementation, QA, and PR creation

Delegate a self-contained task to pi (cheap local AI coding agent) and report back.

Plan a patch release from recent issues. Triages what belongs in a patch, proposes a batch for approval, then opens a tracking issue with the version number, draft release notes and a checklist. Use when asked to plan or scope a patch release, cut a bugfix release, or work out what should go into the next x.y.Z.

Summarise an academic paper from URL, DOI, or file path

Scan repos in an org for standards drift and propagate fixes via worktree PRs

Check which packages in an org are overdue for release and draft release PRs

Set up and run full org automation including repo cloning, script generation, cron scheduling, and skill execution

Clean up stale worktrees and help stuck PRs across an org's repos

Add clarity and structure to open issues across an org's repos without repeating previous work

Audit cross-repo dependencies within an org and flag outdated or breaking changes

Quick scan of CI status across all repos in an org, flagging failures and stale workflows

Create structured meeting minutes from a transcript

Search local bibliography files and Paperpile for relevant papers

Lint, format, and optionally commit changes

Julia package development on a warm REPL session, verifying each edit with Revise and TestItemRunner rather than at the end

Summarise a GitHub issue conversation with focus on the last message

Post a helpful reply to a GitHub issue

Improve test coverage by identifying gaps and generating meaningful tests

Expert knowledge of research grant compliance requirements, deliverables tracking, and funder expectations. Use when reviewing work against grant specifications, preparing progress reports, or ensuring alignment with funding requirements from NSF, NIH, EU, and other agencies.

Expert guidance for setting up and organising research grant applications following established patterns for UKRI, EU, and other funding bodies

GitHub dashboard update with notifications, PR status, issue triage, and action items

Format an Obsidian note with tags and link it from today's daily note

Check that all code has appropriate documentation and fix gaps

Summarise today's Claude Code activity across all repos into a dated log file

Import a markdown file into Obsidian library and format it as a note

View what all active Claude Code sessions are doing and optionally send instructions to them via inbox

Create a well-structured commit with conventional format and proper git identity

Audit and clean up ~/code directory structure, archiving stale repos, fixing folder names, and removing dead worktrees

Check bot notifications for tasks requested by the repo owner, then execute them using a team

Report on what the unattended agent bots have been doing — what the review bot has posted, what it cost, whether anything failed, and whether its findings were acted on. Use when asked "what have the bots been up to", to audit bot activity before trusting it, or to check a bot is still running after a change to its script or cron entry.

Expert methodology for analyzing and summarizing research papers, extracting key contributions, methodological details, and contextualizing findings. Use when reading papers from PDFs, DOIs, or URLs to create structured summaries for researchers.

Add the bot account as a collaborator with push permissions to a repository

Expert knowledge of academic writing standards for peer-reviewed papers, including citation integrity, style compliance, clarity, and scientific writing best practices. Use when reviewing or editing academic manuscripts, papers, or research documentation.

Evaluate Julia through the warm AgentREPL MCP session rather than `julia -e`, hot-reload edits with Revise, and filter TestItemRunner suites while iterating. Use for any Julia evaluation, package iteration, or test run, and to decide when a fresh process is needed instead.