ghostcrab-prompt-guide
Use when a user needs help turning plain-language goals into strong GhostCrab prompts.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Use when a user needs help turning plain-language goals into strong GhostCrab prompts.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Audit gaps between a natural-language GhostCrab question and evidence available through MCP on a SQLite workspace. Use when a projection is missing, a Type A contract exists without evidence, required facets or graph edges are not verified, a Type B snapshot is absent, an MCP tool fails, or the user asks what to add to make an operational answer reliable.
Review GhostCrab projections and explain purpose, readiness, required dimensions, facets, edges, and limits. Use for projection scope review, Type A contracts, Type B snapshots, or readiness on a Personal workspace.
Audit gaps between a natural-language GhostCrab question and evidence available through MCP on a SQLite workspace. Use when a projection is missing, a Type A contract exists without evidence, required facets or graph edges are not verified, a Type B snapshot is absent, an MCP tool fails, or the user asks what to add to make an operational answer reliable.
Review GhostCrab projections and explain purpose, readiness, required dimensions, facets, edges, and limits. Use for projection scope review, Type A contracts, Type B snapshots, or readiness on a Personal workspace.
Transform GhostCrab/MindBrain materialized projection responses into visual, presentation-ready reports with KPI cards, timelines, deltas, heatmaps, evidence tables, provenance, and recommendations. Use when a user wants a projection report to be clearer, more narrative, more visual, or demo-ready.
Audit gaps between a natural-language GhostCrab question and evidence available through MCP on a SQLite workspace. Use when a projection is missing, a Type A contract exists without evidence, required facets or graph edges are not verified, a Type B snapshot is absent, an MCP tool fails, or the user asks what to add to make an operational answer reliable.
SOC 직업 분류 기준
| name | ghostcrab-prompt-guide |
| description | Use when a user needs help turning plain-language goals into strong GhostCrab prompts. |
| disable-model-invocation | true |
Use user language first. Do not expose GhostCrab internals unless the user explicitly asked how the system is implemented.
When the user asks for GhostCrab help and the request is still fuzzy, answer with:
For a first-turn fuzzy onboarding request:
ghostcrab_schema_registerHelp the user discover the shape of the problem without forcing them to know GhostCrab schemas, facets, or graph structure.
If the user asks for a reusable GhostCrab project prompt, ontology setup prompt, or import/mapping prompt, point the agent toward the GhostCrab Personal StarterKit rather than proposing ad hoc files.
Resolve paths first: STARTERKIT_PATHS.md.
For ontology setup prompts, ask the next agent to load:
{starterkit}/personal-mcp/SOP0_import_path_choices.md{starterkit}/personal-mcp/SOP2_obsidian_ontologie.md (§6 bis LinkML default, or §7 MCP Voie A){starterkit}/templates/linkml_ontology.stub.yaml{starterkit}/templates/import_path_choices.yaml{project}/ontology/ for generated LinkML modulesFor tabular source imports, ask the next agent to load:
{starterkit}/personal-mcp/SOP0_import_path_choices.md (§4 tabular choice){starterkit}/personal-mcp/SOP5_structured_import.md (structured-import CLI default){starterkit}/templates/source_profile.yaml{starterkit}/templates/mapping_external_to_canonical.yaml{starterkit}/templates/consumer_contract.yaml{starterkit}/templates/import_manifest.yamlFor multi-module domains with LinkML enums, include ENUM_BUSINESS_FACETS.md in the prompt — facet keys must use <module>.<slot_snake_case> without the user having to ask.
When drafting prompts, always offer both paths as numbered choices; recommend LinkML for ontology and structured-import CLI for tabular on Personal SQLite.