Skill Claude CodeCodex
Specialized debug loop for SnowGram lane/section layout issues. Inherits from snowgram-debugger. Use when: lane badges not appearing, nodes scattered, layout not matching reference PDF. Triggers: debug layout, fix lanes, layout broken, badges missing, streaming diagram, section badges.
Skill Claude CodeCodex
Manage and optimize the SnowGram Cortex Agent for architecture diagram generation. Use when: debugging agent responses, updating agent configuration, adding new components, testing diagram generation, or optimizing the dynamic component system. Triggers: snowgram, diagram agent, component mapping, agent spec, suggest…
Skill Claude CodeCodex
This skill is superseded by snowflake-architecture-diagram, which adds.
Skill Claude CodeCodex
Generalized autonomous debug loop for the entire SnowGram codebase. Use when: any bug, error, test failure, or unexpected behavior. Triggers: debug, fix bug, error, failing test, broken, not working, investigate, diagnose.
Skill Claude CodeCodex
Asks the user 2-4 targeted clarifying questions BEFORE generating a diagram when the request is ambiguous about: data scale, real-time vs batch, governance / compliance constraints, source systems, downstream consumers, or which Snowflake-native services should anchor the architecture. Skip this skill entirely when…
Skill Claude CodeCodex
Surgical edit of an existing canvas diagram. Invoked when the user's request modifies a diagram already present in the conversation — adding components, removing components, renaming, replacing, reconnecting edges, or restructuring sections — instead of generating a brand-new architecture from a template. TRIGGER…
Skill Claude CodeCodex
Generate, view, and export high-quality Snowflake architecture diagrams. Works in two modes: STANDALONE (no Snowflake deployment required, uses bundled templates + local Mermaid composer) and CONNECTED (uses deployed SNOWGRAMAGENT for full conversational generation, or reverse-engineers diagrams from live lineage).…
Skill Claude CodeCodex
Connected-CLI mode: invoke deployed SNOWGRAMAGENT via cortex agents run, parse Mermaid from response, render in local viewer. No live frontend required.
Skill Claude CodeCodex
Connected-UI mode: hand off to the live SnowGram frontend at localhost:3002 for full conversational refinement, multi-tab, per-tab thread state.
Skill Claude CodeCodex
From-lineage mode: reverse-engineer an architecture diagram from real Snowflake objects using cortex lineage. Works in any account — no SnowGram deploy required.
Skill Claude CodeCodex
Standalone mode: generate Snowflake architecture diagrams with no Snowflake connection required. Default path is from-scratch composition informed by SnowflakeProductDocs and SME-skill consultation; the 14 bundled templates are FALLBACKS used only when a prompt cleanly matches a known reference architecture.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: