메뉴
Build the Observable Framework dashboard for a code-diligence target's warehouse, honoring the privacy gate on the working-pattern card. Use when the user invokes /code-diligence:dashboard, or auto-fired by the PostToolUse hook after a warehouse write.
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
Spawn a read-only conversational analyst over a code-diligence target's warehouse and repo clones for open-ended drill-down into a finding (e.g., "why is this repo a hotspot?", "who carries bus-factor risk for X?"). Use when the user invokes /code-diligence:ask. Distinct from investigating which tool fed which warehouse column — use interpreting-tool-output for that.
Initial setup for a new code-diligence target — clone repos, walk git history, run the tool wrappers, populate the warehouse, and generate curation YAMLs. Use when the user invokes /code-diligence:ingest, or sets up code-diligence on a target for the first time (no warehouse exists yet). For already-ingested targets that need new commits picked up, use refreshing-target.
Reference for mapping raw tool output (code-maat, hercules, git-of-theseus, git-fame, git-truck, repowise) to warehouse tables, tool-specific quirks, and cross-check rules. Use when debugging warehouse mismatches, investigating why two tools report different values for the same metric, deciding which canonical source to trust, or querying a specific warehouse table to compare per-tool entries — distinct from open-ended dashboard drill-down (asker) or writing narrative copy (axes).
Incrementally refresh an existing code-diligence target — pull new commits, rerun the tool wrappers, and update the warehouse without resetting analyst-confirmed curations. Use when the target was already ingested (its warehouse exists) and the user invokes /code-diligence:refresh, or asks to recompute, rerun, or update an existing target's data.
Reference for the People / Product / Process taxonomy, card-to-table mappings, confidence model, and narrative-writing contract (headline-first, cite warehouse numbers, hedge on low confidence). Use when interpreting what a dashboard card measures, writing or updating narrative rows, understanding confidence classifications, or avoiding hallucination in PE write-ups — not for querying the warehouse or executing tool commands.
| name | building-dashboard |
| description | Build the Observable Framework dashboard for a code-diligence target's warehouse, honoring the privacy gate on the working-pattern card. Use when the user invokes /code-diligence:dashboard, or auto-fired by the PostToolUse hook after a warehouse write. |
| argument-hint | <target> |
| allowed-tools | Bash, Read |
The user (or the PostToolUse hook) has invoked /code-diligence:dashboard <target>.
Resolve <target> to its settings file at ~/.claude/code-diligence-<target>.local.md. If missing, tell the user and stop.
Read these fields from the settings:
storage.warehouse_path — input DuckDB warehousedashboard.output_path — directory to receive the rendered dist/heuristics.allow_working_pattern_card — privacy gate (default false); see references/privacy-gate.mdBuild the dashboard. Pass --allow-working-pattern only when the flag is true:
uv run --directory ${CLAUDE_PLUGIN_ROOT} \
python -m code_diligence.dashboard.render \
--warehouse <warehouse_path> \
--output <output_path> \
[--allow-working-pattern]
On success, tell the user the output directory and the preview command:
python -m http.server -d <output_path>
On non-zero exit, surface the stderr verbatim. The renderer fails the build when the template's npm install or observable build fails — do not silently retry.
Full renderer argument surface (curation, custom output, env vars): references/render-args.md.
Template layout and what gets copied where: references/template-layout.md.