Reusable skills for AI coding agents.
midjourney-images writes focused Midjourney prompts and can create images in the Midjourney web app. Use it to explore a visual style, build a brand world, or develop concept art.
The skill supports two tasks:
- Prompt only: Write or improve a prompt without opening Midjourney.
- Create on web: Submit one standard image batch and report the result.
Install it with the Skills CLI:
npx skills add transitive-bullshit/skills --skill midjourney-imagesbranding-exercise develops a project brand through a short context interview, distinct visual identity directions, and iteration. Use it to create a new brand identity or intentionally rebrand an existing project.
The skill produces:
- Visual directions: At least three distinct brand concepts with one-page posters to compare and refine.
- Accepted identity: A brand guide, reusable logo and imagery assets, favicons, a social image, and a reproducible one-page poster.
- Project guidance: A pointer in the project's agent instructions so future customer-facing work follows the accepted identity.
Install it with the Skills CLI:
npx skills add transitive-bullshit/skills --skill branding-exercisestorytelling creates and critiques marketing narratives, product or project pitches, and personal stories. Use it to sharpen a hook, clarify emotional stakes, connect story beats, or make an ending land.
This skill is based on work by Edmund Tian.
The skill supports two tasks:
- Create: Draft a narrative around a clear message, causal story beats, and a memorable payoff while preserving your voice and grounding claims in supplied facts.
- Critique: Identify the changes that most improve clarity, stakes, credibility, or payoff, with concrete revisions to weak passages.
Install it with the Skills CLI:
npx skills add transitive-bullshit/skills --skill storytellingx-data sets my X lookup order: local Birdclaw → authenticated bird → public FxTwitter → paid xurl. The archive and API skills and the posting routines follow this shared policy.
social-posting and x-mirror are my own social posting setup: launch game plans and scheduled posts across Threads, Bluesky, LinkedIn, Instagram and TikTok through Postiz, using my X posts as the guide. They live in routines/ because only my scheduled Claude routines use them, so they aren't installed as skills. They're wired to my accounts and here as a reference.
Two more follow my own conventions: init-git-repo sets up a clean, shareable repo the way I like it (audit, baseline files, my TypeScript tooling, docs), and personal-notion-projects drafts project write-ups in the Notion CMS behind my site.
For a shared setup across Codex and Claude, give your agent this prompt:
Set up transitive-bullshit/skills for my coding agents. Read AGENTS.md,
inspect existing skills, and use the mac profile for a fresh installation.
Preserve existing profiles and local edits, review conflicts before adopting
them, and finish with the doctor check.
From the repository, with Python 3.11+:
python3 scripts/skillset.py apply
python3 scripts/skillset.py doctorThis reuses your installed profile (or defaults to mac) and symlinks skills into ~/.agents/skills and ~/.claude/skills. Edits through either path are Git changes here. Commit and push normally; pull and apply on other machines. Existing files are preserved; conflicts stop installation.
See setup notes for profiles and adoption, or the dotfiles repo for full environment setup and agent-env synchronization.
skills.json records every source, pinned revision, license, profile, and local patch. Pristine upstream resources and reconstructable patches live in provenance/; the pre-curation collection is preserved in history.
Run python3 scripts/upstream.py check to verify the imported sources plus local patches reproduce the installed skills. See setup notes for the update workflow.
Personal skills are MIT by Travis Fischer. Third-party skills retain their upstream licenses and attribution; available license files and unresolved provenance are recorded in skills.json.