Agent Claude Code
Given ONE Sentry issue short-id, investigates the root cause and — in active mode — opens a DRAFT PR with a minimal, tested fix; in report-only mode (the default) it proposes the fix without pushing; or it DECLINES with a reason. Never auto-merges, never touches main, never edits auth/security/payments/migrations.…
Agent Claude Code
Nightly Sentry orchestrator for the Shelf webapp. Classifies every unresolved issue; resolves noise/transient/already-fixed in Sentry with a reason; dispatches a sentry-issue-fixer per genuine, localized, auto-fixable bug; flags the rest for a human; emits a morning digest. Defaults to report-only. Use for "triage…
Agent Claude Code
Verifies a single PR review finding from CodeRabbit, Codex, or Copilot against the current state of the Shelf codebase and returns a structured verdict. Read-only — never edits code and never writes to GitHub. Invoked by the /pr-review-loop skill, one instance per finding, in parallel.
Agent Claude Code
Headless variant of shelf-security-reviewer for the pre-commit hook. Receives the staged diff inline (the wrapper script pre-computes it). Tools restricted to Skill only — no Bash, no WebFetch, no Agent — so prompt injection in the diff cannot exfiltrate. Outputs a strict JSON envelope for safe parsing. Do NOT invoke…
Agent Claude Code
Security-focused reviewer for Shelf PRs and diffs. Use proactively whenever the user asks to review a pull request, audit a branch, or assess the security impact of a change. Always invokes the security-review, oauth-oidc-misconfiguration, supabase, and two-factor-authentication-best-practices skills, and layers…
Command Claude Code
Nightly Sentry triage — classify the board, propose or open tested fixes, and print the morning digest. Report-only unless you pass "active".
Hook Claude Code
Runs when the agent finishes a response, running afplay ~/Documents/jobs_done.aiff. From Shelf-nu/shelf.nu.
Hook Claude Code
Runs after a tool call finishes for Edit, Write and MultiEdit tool calls, running npx prettier --write "$FILE_PATH". From Shelf-nu/shelf.nu.
Settings file Claude Code
Agent settings declaring 2 hook events (Stop, PostToolUse) and 37 allowed tools.
Skill Claude CodeCodex
Automate the PR review-response cycle — watch a PR for CodeRabbit/Codex/Copilot/human feedback, verify each finding, implement the valid ones, commit, then reply and resolve on GitHub. Iterates until the user stops it. Use when the user says "start the PR loop", "handle the review comments", or invokes /pr-review-loop.
Skill Claude CodeCodex
Use when finishing a feature, fixing a bug, before committing React code, or when the user wants to improve code quality or clean up a codebase. Checks for score regression. Covers lint, dead code, accessibility, bundle size, architecture diagnostics.
Skill Claude CodeCodex
Use when formatting a GitHub release for Shelf — turns GitHub's auto-generated "What's Changed" list into the house format (categorised sections, highlights, security block). Invoke for "format the release notes", "write the release notes for X", or when preparing a draft release.
Instructions file GitHub Copilot
Copilot instructions for Shelf-nu/shelf.nu, covering github copilot instructions for shelf.nu, repository overview, working effectively, prerequisites and bootstrap and setup commands.
Instructions file CodexOpenCode
AGENTS.md instructions for Shelf-nu/shelf.nu, covering agents.md, monorepo structure, key commands, development and database.
Instructions file
Claude Code instructions for Shelf-nu/shelf.nu, covering claude.md, essential commands, webapp, companion app (mobile) and docs.