mdelapenya

Perform a rock-solid, multi-pass adversarial review of a pull request. Works from inside the matching repo (bare PR number, infers org/repo from git remote) or from a barebones directory (PR URL or org/repo#num shorthand). Clones the PR head into a temp dir, or uses the current branch when it is already the PR. Runs an adversarial first pass, then re-evaluates and challenges its own findings to drop false positives and confirm real ones. Designed to be invoked iteratively in a loop until findings converge. Use when user says "review this PR", "adversarial review", "deep PR review", "challenge this PR", "audit this pull request", "stress test the PR", or "pr reviewer".

Generate a concise PR description and a Conventional Commits title from a GitHub pull request diff. Auto-detects remotes, honors PR templates, and updates both title and body via gh. Use when user says "describe this PR", "add PR description", "fill in PR body", "describe pull request", "summarize PR changes", or "update the PR title".

Investigate CI failures on a PR by cross-referencing failing tests against other recent runs to determine if failures are pre-existing or caused by the PR's changes. Use when user says "investigate CI failure", "is this test flaky", "why is this test failing", "cross-reference builds", "check if failure is pre-existing", or "ci detective".

Monitor CI builds and mergeable status for the current branch's PR. Diagnoses failures, merges main to resolve conflicts or pick up fixes, and pushes. Use when user says "nurse this PR", "check CI", "fix the build", "why is CI failing", "merge main", or "resolve conflicts".

Review and respond to PR comments by either fixing the code (when feedback is valid) or challenging the reviewer with a reasoned argument (when feedback is debatable or wrong). Use when user says "address PR comments", "respond to review", "challenge this comment", "handle review feedback", "fight back on review", or "pr lawyer".

Generate social media content for LinkedIn and Twitter/X to promote blog posts from your Hugo blog. Reads a post, extracts key insights, and produces platform-specific drafts. Use when user says "promote this post", "share on LinkedIn", "tweet about this", "social media for this post", or "market this article".

Create new blog posts for your Hugo blog. Handles front matter, structure, tone, and drafting through an interview-driven workflow. Use when user says "create a new post about", "write a post", "draft a post", "new blog post", or asks to write blog content from scratch. Do NOT use for editing or reviewing existing posts.

Audit internal links, cross-references, and "coming soon" teasers across all blog posts on your Hugo blog. Reports broken links, unfulfilled teasers, missing bidirectional series references, and optionally checks external URLs. Use when user says "check blog links", "audit posts", "find broken links", "blog-keeper", or "validate cross-references". Read-only — does not edit files.

Discover and plan new blog post topics for your Hugo blog. Analyzes existing posts to suggest expansion topics, then interviews the author to shape the angle. Use when user says "new post idea", "what should I write about", "suggest a topic", "brainstorm posts", or "plan a new article".

Review and edit existing blog posts for your Hugo blog. Enforces front matter conventions, tone, structure, and conciseness. Use when user says "review this post", "edit the blog post", "check my front matter", "trim this section", "review my post", or asks about blog editing style. Do NOT use for creating new posts (use blog-writer) or general writing tasks unrelated to blog posts.

Functional knowledge base for the biomelab product — user-facing features, workflows, and decision log. Use when discussing product requirements, planning new features, evaluating UX changes, writing user-facing copy, triaging bugs from a user perspective, or answering "what does biomelab do?" questions. Do NOT use for implementation/code questions.

Technical reference for building desktop GUI applications with Fyne (v2.7+). Use when implementing UI features, debugging rendering/layout issues, handling keyboard/mouse input, creating custom widgets, or working with dialogs and themes. Do NOT use for business logic or non-GUI questions.