AI Agent Workflow Collection

Create, plan, direct, critique, generate, or edit slow-travel and lifestyle photography using the visual language associated with Italian photographer Roberta Mazzone (@lagiuditta): architectural composition, natural light, warm restraint, lived-in detail, and quiet cinematic narrative. Use when the user names Roberta Mazzone or @lagiuditta, asks for this photographic direction, or requests an adjacent Mediterranean travel, boutique-hotel, architectural, lifestyle, portrait, food, or still-life image treatment.

Research, art-direct, create, and critique original alternate movie, television, game, and pop-culture posters using the curatorial and screen-print production principles associated with Mondo: deep source knowledge, a fresh conceptual hook, artist-to-property fit, integrated typography, collectible-object craft, variants, and rigorous review. Use when a user mentions Mondo, Mondo式/Mondo风格海报, alternative movie posters, collectible screenprints, licensed pop-culture key art, or wants a poster brief, concept routes, generation prompt, critique, or print plan in this tradition.

Generate and integrate exactly one article preview image for cankun-blog by inferring one visual metaphor from a target MDX post, maintaining the Franklin Booth-inspired neutral-white series, updating the prompt manifest, creating the 2000×800 PNG original, running the repository optimizer for the 1400×560 WebP, wiring bilingual alt text and the full-resolution link, validating, and optionally publishing. Use when the user asks to add, make, generate, refresh, replace, or remake a preview, hero, or cover image for one existing or newly added cankun-blog article.

Consult a peer AI model (Claude or Codex, plus GPT Pro through Oracle when explicitly requested) and keep the dialogue resumable across rounds. Use when the user wants a second opinion from another model ("ask codex", "问问 codex", "让 claude 看看", "gpt says..."), explicitly asks to use/ask GPT Pro or Oracle, wants a multi-round cross-model review, wants to resume or continue an earlier consultation thread, or when a conclusion needs independent-model verification.

Depose a plan, decision, or idea — relentless one-question-at-a-time examination with a visible record (✓ settled · ? open · ~ unwalked) and a nothing-further closing gate. Use when the user wants to stress-test their thinking before acting, says "depose this", or uses any 'grill' trigger phrases.

Orient a multi-turn “help me think this through” deliberation with a Rumsfeld matrix. Use when the user owns an open decision or judgment and wants a thinking partner — articulating a tacit judgment, surfacing an unrecognized assumption or pattern, or touring options around the choice they are weighing. A question seeking a direct answer (factual lookup, how-to, debugging) gets the answer, not this skill.

Wayfind repository documentation by tracing real tasks to code-owned authority. Use when the user wants to audit scattered, stale, duplicated, or mixed-role docs; decide whether README, AGENTS.md, architecture, runbook, or protocol documents should exist and where they belong; or apply an accepted documentation-topology repair.

Release and sync skills in the cankun-skills collection. Use when the user wants to publish or release a skill, add a new skill to the index, sync or fix a drifted skills README or the cankun.me skills page, run publish-hygiene checks, or promote a skill to its own flagship repo (or fold one back).

从旅行社行程 PDF + 机票酒店票据生成 A5 手机优先的自足旅行手册(瑞士极简…