Data & Analytics

Use when a pipeline fails because source, destination, or declared asset columns may have changed.

Use when a Bruin quality check fails or starts warning unexpectedly.

Use when a Bruin pipeline, asset, or command fails and the cause is not yet clear.

Use after a diagnosis skill has identified a likely fix and the repository owner wants to define a controlled action.

Use when data is stale, a scheduled run is missing, or freshness checks fail.

Use when duplicate rows, unstable primary keys, repeated ingestion, or failed uniqueness checks appear in a Bruin asset.

Use when creating, editing, reviewing, or troubleshooting Bruin semantic layer models, semantic query CLI usage, metric and dimension definitions, joins, segments, filters, windows, or semantic-layer tests and docs in a Bruin repository.

Run an extremely strict maintainability review for abstraction quality, giant files, and spaghetti-condition growth. Use for a thermo-nuclear code quality review, thermonuclear review, deep code quality audit, or especially harsh maintainability review.

Create, update, render, and visually verify polished Bruin CLI terminal demos with VHS. Use when recording a terminal walkthrough, adding or editing a .tape file, exporting a demo video or GIF, improving terminal-demo styling or readability, or reproducing a Bruin command sequence for documentation or social media.

Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, passive voice, negative parallelisms, and filler phrases.

Add Bruin CLI support for a new ingestr source. Use when a task asks to implement a new ingestr connection/source type, wire that source into Bruin ingestr assets, update available source tables, add ingestion docs or example assets, regenerate connection schema expectations, or address review feedback for a newly added ingestr source.

**Role**: Step 0 — Evaluate whether existing snapshot data is fresh enough to reuse, or whether new data collection is required. **Triggered by**: CLAUDE.md at the start of Workflow 1 (Single Stock Analysis) and Workflow 3 (Watchlist Scan) **Reads**: `output/data/{ticker}/latest.json`, staleness rules from this file **Writes**: Nothing (read-only evaluation; result is reported inline to orchestrator) **References**: `references/staleness-rules.md` ---

**Role**: Step 1 — Parse the user's query to extract ticker(s), determine output mode, detect language, and validate intent. **Triggered by**: CLAUDE.md after Step 0 determines fresh collection needed **Reads**: User query, `references/ticker-resolution-guide.md` **Writes**: Sets session variables: ticker, market, output_mode, output_language, peers (if multi-ticker) **References**: `ticker-resolution-guide.md` ---

**Role**: Step 9 — Perform output-facing quality checks and rebuild the deterministic run-local quality report before delivery to user. Auto-patch minor issues; flag persistent failures inline. **Triggered by**: CLAUDE.md after Step 8 (output generation), before final delivery **Reads**: Generated output file (or inline response), run-local `validated-data.json`, run-local `evidence-pack.json`, optional run-local `context-budget.json`, run-local `analysis-result.json` **Writes**: Patches to t...

**Role**: Step 8 — Generate the final analysis output in the requested mode (A, B, C, D). Mode A delegates to briefing-generator; Mode B uses `scripts/render-comparison.py`; Mode C delegates to dashboard-generator; Mode D uses `scripts/docx-generator.py`. **Triggered by**: CLAUDE.md after Analyst Agent (Step 7) completes **Reads**: `output/runs/{run_id}/{ticker}/analysis-result.json`, appropriate mode template **Writes**: File (Mode B, D); Mode A delegates to briefing-generator/SKILL.md; Mode...

**Role**: Step 2 — Detect MCP availability, determine data mode (Enhanced/Standard), classify company type, identify peers, initialize a run-local artifact root, and write the research plan. **Triggered by**: CLAUDE.md after Step 1 (query interpretation) **Reads**: Session state from Step 1, `references/company-type-classification.md` **Writes**: `output/runs/{run_id}/{ticker}/research-plan.json` **References**: `company-type-classification.md` ---

**Role**: Step 10 (post-analysis persistence) + Workflow 3 (portfolio & watchlist management) **Triggered by**: CLAUDE.md after Step 9 (quality check) for persistence; directly for Workflow 3 commands **Reads**: run-local `analysis-result.json`, optional run-local `evidence-pack.json`, optional run-local `context-budget.json`, `output/watchlist.json`, `output/portfolio.json` **Writes**: Snapshot files, `output/watchlist.json`, `output/portfolio.json`, `output/catalyst-calendar.json` **Referen...

**Role**: Step 8 — Generate the Mode C HTML dashboard from `analysis-result.json`. **Triggered by**: CLAUDE.md when `output_mode = "C"` after Step 7 (Analyst Agent completes analysis) **Reads**: run-local `analysis-result.json`, `references/html-template.md`, `references/color-system.md` **Writes**: `output/reports/{ticker}_C_{lang}_{YYYY-MM-DD}.html` **References**: `html-template.md`, `color-system.md`, `docs/adr/0001-mode-c-rendering-strategy.ko.md` ---

**Role**: Step 8 (Mode A) — Generate the Mode A Quick Briefing HTML file from analysis-result.json. **Triggered by**: CLAUDE.md after Analyst Agent (Step 7) completes for Mode A **Reads**: run-local `analysis-result.json` **Writes**: `output/reports/{ticker}_A_{lang}_{YYYY-MM-DD}.html` **References**: `references/analysis-framework-briefing.md`, `references/html-template.md` (this directory), `scripts/render-briefing.py` ---

Reconciles known (label, expected_value) numbers from an existing dashboard against agami's answers. Input can be a SCREENSHOT of a Metabase / Power BI / Tableau / Looker dashboard (Claude's vision extracts the pairs), a CSV, or numbers pasted inline — the user doesn't need to know which; they can just ask. For each pair, the skill generates a matching NL question, runs it through the active profile's semantic model, diffs actual vs expected, and surfaces matches in green and mismatches in re...

For each new or changed chunk in chunks/ (by git), find or create a GitHub issue for its PMID, then commit the chunk with "Fixes #issue_id". Processes one chunk at a time, always asking user before creating issues or committing.

Run QC scripts on a VDJdb chunk, report every error with a suggested fix, verify the output of previous /extract and /format steps, estimate confidence scores, and flag gaps in current py_src QC coverage.

Harmonize antigen.gene and antigen.species fields in a VDJdb chunk to canonical VDJdb naming. Detects spurious gene/species names, resolves inconsistencies (same epitope → multiple names), and warns about epitopes that are exact substrings of longer epitopes. Invoked standalone or from /proofread when spurious values are detected.

Standardise a raw or partially-formatted VDJdb TSV chunk — normalising species names, IMGT V/D/J gene IDs, IMGT-HLA MHC alleles, and method vocabulary — and produce a properly-named chunk file ready for proofreading.