메뉴
Searches the format-database/ KDB for a schema mapping that matches the input merger tree format. Use when a new input file has been identified and needs to be matched against known formats before any conversion begins.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | kdb-lookup |
| description | Searches the format-database/ KDB for a schema mapping that matches the input merger tree format. Use when a new input file has been identified and needs to be matched against known formats before any conversion begins. |
Output the following to the user verbatim before any other action:
Stage 1 — Discovery
I'll identify your input format and map its fields to the SAGE LHaloTree schema.
1. Inspect input files
2. KDB lookup
- match -> load schema mapping
- no match -> web discovery + schema mapping
3. [G1] Confirm mapping + select output format
.ai/skills/kdb-lookup/references/<file> — files in this skill's own references/ subfolder.format-database/ — the project-root format database directory.Before searching the KDB, identify the file type. Apply the file inspection rule: read only the minimum needed.
head -n 30 <file> — record the comment character, column headers, and first data rows.h5dump -n <file> — record the group and dataset structure only. Do not read dataset values unless a small sample is needed to identify a field (h5dump -d <dataset> --start="[0]" --count="[5]").xxd <file> | head -4 — record the first 64 bytes (magic bytes and header fields).From the inspection, determine:
If any of the three identifiers are uncertain, record them as uncertain and note what additional information is needed.
Read every .json file in format-database/. For each entry, compare:
halo_finder against the identified halo findertree_tool against the identified tree toolfile_format against the identified file formatA full match requires all three to agree (case-insensitive comparison is acceptable). Do not accept a two-out-of-three match silently.
If a partial match is found (two of three agree), flag it explicitly to the user:
Partial match found: format-database/<entry>.json
halo_finder: <value> — MATCH / MISMATCH
tree_tool: <value> — MATCH / MISMATCH
file_format: <value> — MATCH / MISMATCH
I am treating this as NO MATCH. Please confirm or provide the missing identifier.
Wait for the user's response. If the user supplies the missing identifier and all three criteria now agree, proceed with Step 3 (full match path). If the user cannot confirm or the mismatch remains, proceed with Step 4 (no-match path, invoke web-discovery).
See .ai/skills/kdb-lookup/references/matching_criteria.md for the full match definition and edge cases.
If a full match is found:
format_id, description, driver_module, and all known_caveats.field_map as a readable table: SAGE field → source field → units → conversion expression.pointer_logic section.AGENTS.md).Do not proceed to Stage 2 until the user confirms.
After the user confirms (G1 passed), copy the matched KDB entry to
assets/proposed_mapping_<format_id>.json:
cp format-database/<format_id>.json assets/proposed_mapping_<format_id>.json
This satisfies the Stage 2 entry condition (AGENTS.md §4) and ensures the confirmed mapping is available for Stage 4 archiving.
If no full match is found after scanning all entries in format-database/:
No matching KDB entry found for the identified format:
halo_finder: <value>
tree_tool: <value>
file_format: <value>
Proceeding to web discovery. I will now invoke the web-discovery skill
to search for external documentation on this format.
Then invoke the web-discovery skill.
Guides drafting or adapting a format-specific conversion driver for the SAGE merger tree converter. Use when no existing driver matches the input format and a new driver must be written in assets/ before the test conversion can run.
Provides field-level guidance for mapping input merger tree fields to the SAGE LHaloTree schema (HDF5 or binary). Use during web discovery or KDB entry creation when individual field mappings, unit conversions, or pointer reconstruction decisions need to be made.
Independently verifies the semantic validation plots and plotting code against the mandatory checklist. Use after semantic-validation generates the seven plots, before presenting results to the user. Operates independently - does not rely on the main agent's assessment.
Generates the seven mandatory semantic validation plots of the converted output merger trees. Use after a full conversion run in Stage 3 to assess whether the converted trees are physically plausible.
Runs the six structural checks on a converted SAGE LHaloTree HDF5 file. Use after any conversion run — test or full — to verify the output file is structurally correct before proceeding.
Adds a new format driver and schema mapping JSON to the KDB after a successful conversion of a previously unknown format. Use in Stage 4 when the input format was not found in the KDB at session start.