Claude Skills by ozlar34

Develop a fleeting idea through adaptive one-question-at-a-time dialogue — figure out what it actually is, what could be done with it, and what the next step would be. User-invocable only (via /braindump); does not auto-trigger. Distinct from an adversarial plan stress-test and from a build-this-now spec session — braindump stays open-ended and writes nothing.

Process an Apple Reminders capture list one item at a time — ideas captured via Siri on mobile. Analyze each against current context, recommend routing (act/save/backlog/ignore), execute on confirmation, then mark complete. Configure the list name and routing destinations below.

Polish a public GitHub repo's presentation — sharpen topics + description, reframe the README as a worked example, add a LICENSE if missing, commit and push every CLI-doable change, and hand back a GitHub-UI checklist for what the API can't do (pin order, social-preview upload, real screenshots). Not for adding features — for how the repo reads to anyone who opens it. Use whenever the user says "/github-polish <repo>", "polish / spruce up / clean up <repo>", "the README is weak", or points at...

Run at the end of a session to extract wrong-mental-model failures (not inefficiencies) and propose surgical, cost-justified CLAUDE.md additions that prevent the same wrong action in future sessions. User-invocable only (via /learn-from-mistakes); does not auto-trigger. Distinct from /session-handoff (which captures work state) — this skill captures behavioral corrections.

Get a fresh, independent Opus agent to verify the last load-bearing claim the assistant made — confirmed / refuted / can't tell, with evidence. On-demand capability escalation for cheaper sessions (e.g. Sonnet) that hit a moment where a claim actually matters. User-invocable only (via /second-opinion); does not auto-trigger. This audits the truth of a stated claim — not whether an app runs.

List an item on Kleinanzeigen. Gathers item details conversationally, writes to <ITEMS_DIR>/current.yaml, then drives the Kleinanzeigen post-ad form via Playwright MCP, narrating each field in English so a non-German-speaker can verify. Stops before publish for manual review. Use when user says "/sell", "list on Kleinanzeigen", "list this on KA", or mentions selling something.

Generate a structured mid-session handoff block. Invoke proactively when remaining context drops to ~15% and meaningful task state would otherwise be lost; also on explicit /session-handoff. Persists to ~/.claude/handoffs/ and copies to clipboard so a fresh session can resume without losing state. Supports `--minimal` for a ~80-token bare-bones block.

Triage items sitting in the canonical Inbox/ (web-clipper, Raindrop, Telegram, manual). THREE modes. Dedicated commands: `/triage-propose` (Mode A — async analysis, stages proposals, Telegram push) and `/triage-review` (Mode B — apply staged proposals). Bare `/triage` runs Mode C — the legacy per-clipping interactive walkthrough, bounded to 2 actionable verdicts/session. If .triage/pending.json exists when running bare, offers to switch to review mode instead. Atomicity is promote→flip→destru...