caffeinity-skills

Everything in one install: every caffeinity skill plus twelve workflow agents (skill-builder, discovery-lead, release-gatekeeper, seo-lead, job-search-coach, decision-analyst, learning-coach, outbound-lead, solo-launcher, skill-maintainer, estate-lead, harness-lead).

60 skills · 12 agents

Audits AI agents and Claude Code subagents against the seven layers a production agent needs: model, instructions, tools, memory, loop, guardrails and human approval, with a linter and delegation tests.

1 skill

Produces tailored job-application material: decoded JD table, action-task-result-impact bullet rewrites, honest keyword placement map, five role-specific summaries, full LinkedIn rewrite, and a 20-question interview answer pack. Sibling of recruiter-lens: it produces while recruiter-lens judges. Never invents experience or numbers.

1 skill

Runs the job search around the applications: search brief, role discovery with honest posting dates, rubric-based fit scoring, a reviewable CSV or Markdown tracker, follow-up drafts, interview briefings and a funnel review. Never submits, invents nothing, and has no CLI, telemetry, or browser automation. Coordinates with applicant-kit and recruiter-lens.

1 skill

Start here: picks which caffeinity skills to run, in what order, for a goal that spans several of them, and names the ones you have not installed.

1 skill

Gates a cold-email lead list before it loads: samples ten rows against the ICP, returns GO or HOLD with the fail pattern and the re-run condition, and compares data providers on the same criteria. Not for lists that already sent; that diagnosis is cold-email-ops.

1 skill

Plans and audits a cold-email program at volume: inbox and domain capacity from a monthly send target, reply-rate floors by domain, burn replacement, shadow-ban tests, automation error budgets and per-client reporting. Thresholds stay labelled as source benchmarks, and missing data returns INCONCLUSIVE rather than a pass. Not for pre-load list QA; that is cold-email-list-gate.

1 skill

Designs a course for a class, cohort or workshop: measurable outcomes, modules in prerequisite order, aligned assessments and a time budget, checked by a script.

1 skill

Crash-tests a decision before you build it. Four fixed roles (Pragmatist, Skeptic, Specialist, Contrarian) draft the same task in separate contexts, attack each other's drafts anonymously, and a chair merges what survived, keeps a dissent log, and grounds the winner's claims against your source before adopting it. Optional Operator seat for cost, rollout and reversibility. No server, no API keys, no other agent CLIs.

1 skill

When an answer is bad, runs a single-elimination tournament of sub-agents with different strategy cards until one solution survives.

1 skill

Creates a new agent skill from a plain idea. Confirms the purpose before writing anything, checks the already-installed siblings for overlap, splits the work between scripts, references and assets, writes a description that actually triggers, then validates with tune-skill's check_skill.py and hands off for the boundary check. Original authoring skill: it contains no text from Qoder's built-in skill-creator, which ships without a licence permitting redistribution.

1 skill

Sets up discovery for a GitHub repo, agent skill, Claude Code plugin or MCP server so people and agents find it by what it does, not just its name: audits current visibility and same-name collisions, derives the queries strangers actually type, drafts the About text, topics, README opening, release notes and a project page, then plans directories and a dated re-check. Never games stars or leaves example slots with invented output.

1 skill

Plans topic clusters and kills keyword cannibalization so exactly one page wins each query, with a mandatory cannibalization audit before content changes.

1 skill

Drafts the privacy policy, terms, refund policy and cookie notice a small product needs from a confirmed fact sheet, and checks them for placeholders, missing sections and invented facts. Not legal advice.

1 skill

Audits and fixes web app performance before launch: four read-only subagent audits over twenty known slow-downs across images, fonts, bundles, database queries and servers, then same-condition mobile Lighthouse medians prove each fix. Never removes a feature for a score; asks before dependency, index, RLS or region changes.

1 skill

Authorized, evidence-driven red-team pass over a web app you own, before launch. Three read-only subagents run 14 authorization and abuse checks against isolated synthetic accounts, the chair fixes each failure, adds regression tests, and reruns with a negative control. Isolated local environments only; secrets masked in every artifact; anything unrun is reported not-proven, never passed.

1 skill

Turns a list of your own prompts into a finished, tested agent skill, one skill per prompt. Interviews for intent, checks what is already installed so the new skill does not duplicate it, drafts with draft-skill, writes trigger test cases including near-misses, then gates the result with tune-skill. Requires draft-skill and tune-skill and stops with a clear message when either is missing. Does not do the release packaging.

1 skill

Packages a finished agent skill for release: runs its trigger test cases and records exactly which passed and which did not, sets the version and CHANGELOG, builds the archive, writes the README, and ends with a per-platform install checklist. Failed tests stay visible in the report instead of being dropped from it. Assumes the skill is already authored, by prompts-to-skill or draft-skill.

1 skill

Owns the machine that turns skill repos into a published site: one committed manifest as the publication decision, discovery that opens a pull request, a generated page per new skill, both discovery indexes, a sitemap, and gates that fail closed before and after the deploy. Not for releasing one skill through the marketplace or tarball channels; that is publish-release-gate.

1 skill

Publishes a skill through three distribution channels: marketplace plugin, agent discovery protocol, and tarball delivery via Cloudflare Pages custom domain. Use when someone says 'publish to marketplace', 'make discoverable by agents', 'deploy skill site', or after completing a skill update that needs public distribution.

1 skill

Reviews resumes, CVs and LinkedIn profiles the way recruiters screen them: first-impression scans, ATS compatibility, rejection-pattern diagnosis and interview prep. Never invents a metric.

1 skill

Runs three pre-launch passes (exposure, app-tests, perf) in sequence and returns one release posture — HOLD or ready — with three separate verdicts and one merged finding register. Deterministic arithmetic over union of findings; a lane that did not run holds the gate. Use when asked 'is this safe to ship', 'ready to deploy', or for go/no-go decisions before any production release.

1 skill

Runs web-preflight, the security red team and the performance audit in order, then computes one release posture (HOLD or ready) with a script. Three verdicts stay separate.

1 skill

Crawls a site you own and returns deterministic, evidence-backed technical SEO findings: indexability, canonicals, titles, metas, H1s, broken internal links, sitemap coverage and static JSON-LD. Bounded and read-only; blocked or unrun checks are reported INCONCLUSIVE, never a pass. Feeds seo-audit and keyword-guard.

1 skill

Takes a whole collection of skill repos from scattered to published and self-maintaining, in gated phases handed to the skills that own them.

1 skill

A world-class teacher for any topic, from zero background toward expert depth, in eight modes from foundations and curricula to strict exams and ELI5.

1 skill

Takes one published skill at a time from as-is to tuned, fact-checked, released and live: tune-skill, a fact check against primary sources, findable-polish, one PR, then a confirmed refresh of the site and marketplace copy.

1 skill

Audits and fine-tunes an agent skill that already exists: the description that never fires, overlap with installed siblings, stale names and counts, and handoff contracts that do not hand off. Ships check_skill.py, a deterministic SKILL.md validator covering frontmatter, folder and heading agreement, description length, reference paths that do not exist, and count phrases that go stale. Not for creating a skill from an idea; that is draft-skill.

1 skill

Audits local git clones against their upstream before you push. Fetches first, classifies each repo as level, push-ready, behind or diverged, screens the unpushed diff for secrets, and only pushes with an explicit --push. A second script compares your published listings against the repos that actually ship a SKILL.md. Never force-pushes; a behind or diverged repo is reported, not touched.

1 skill

Checks a site you own before you ship it, treating status codes as signals rather than findings; unrun checks are reported as INCONCLUSIVE, never a pass.

1 skill

Teaches your agent to edit, not just generate: 44 commands across 5 disciplines that tighten, rewrite, critique and proofread while keeping the author's meaning, hedges and voice intact.

1 skill

Mod: reminds Claude to run the pre-ship performance audit before a production deploy. Bundles pre-ship-performance-audit.

Mod: holds production deploys until the pre-ship security red team has run since the last edit. Bundles pre-ship-security-red-team.

Mod: holds production deploys, PR merges, pushes and publishes until the release-gate chair has run since the last edit. Bundles release-gate-chair.

Mod: holds gh release create, npm publish and version tags until prompts-to-skill-release has run its test cases. Bundles prompts-to-skill-release.

Mod: after Claude edits a SKILL.md, nudges it to validate with tune-skill before pushing or releasing. Bundles tune-skill.

Mod: reminds Claude to run web-preflight before a production deploy (wrangler, vercel --prod, netlify --prod, firebase, fly). Bundles web-preflight.