# Personal Finance Update

> Run the finance pipeline, extract numbers from master CSVs, diff against Obsidian note, and update stale tables. Use when user says '/personal-finance-update'.

- Skill: `hosaypeng/personal-finance-update` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add hosaypeng/personal-finance-update`
- Raw SKILL.md: https://api.skillmd.com/api/skills/hosaypeng/personal-finance-update/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: DevOps & Infra
- Author: hosaypeng (https://skillmd.com/u/hosaypeng)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/hosaypeng/personal-finance-update

---


# Finance Pipeline Refresh

Run the full finance pipeline, extract current numbers from master CSVs, diff them against the Obsidian personal finance note, and update any stale tables in-place.

---

## Phase 0: Pre-flight

1. Read `~/.claude/skills/personal-finance-update/config.json`.
2. Validate every value path exists on disk.
3. If any path is missing: **STOP.** Report `MISSING: <key>: <path>` to the user and do not proceed. Do not attempt to infer a new location or continue with partial data.

---

## Phase 1: Pipeline

1. Run the finance pipeline:
   ```bash
   bash the `pipeline_script` path from config.json --quick
   ```
2. Check the exit code:
   - **Exit 0**: Continue silently to Phase 2.
   - **Exit 1**: Warn the user about validation warnings, then continue to Phase 2.
   - **Exit 2**: STOP. Report validation failures to the user and do not proceed.

---

## Phase 2: Extract

1. Run the extraction script:
   ```bash
   python3 ~/.claude/skills/personal-finance-update/scripts/finance_extract.py
   ```
2. Check the exit code:
   - **Exit 0**: Capture the JSON output from stdout. Save it to `/tmp/finance_extract_output.json`.
   - **Non-zero**: **STOP.** Report the stderr output to the user and do not proceed.

---

## Phase 3: Diff

1. Run the diff script against the Obsidian note:
   ```bash
   python3 ~/.claude/skills/personal-finance-update/scripts/finance_diff.py /tmp/finance_extract_output.json the `obsidian_note` path from config.json
   ```
2. Capture the JSON output. Also check stderr — if any `Warning: section '...' not found` messages appear, note the missing sections and report them in Phase 7.
3. If `cells_changed` is 0: report "All tables up to date -- 0 changes" and skip directly to Phase 6.

---

## Phase 4: Update

For each table in `tables_changed`:

1. Use the Edit tool to replace the section in the Obsidian note.
2. The `old_string` is everything from the `heading` line through the end of the `postscript`.
3. The `new_string` is the regenerated heading + preamble + table + postscript from `new_markdown_sections`.
4. Only touch tables that actually changed. Leave unchanged tables alone.

---

## Phase 5: Verify

1. Run `finance_extract.py` again:
   ```bash
   python3 ~/.claude/skills/personal-finance-update/scripts/finance_extract.py
   ```
2. Save to a new temp file and run the diff again:
   ```bash
   python3 ~/.claude/skills/personal-finance-update/scripts/finance_diff.py /tmp/finance_extract_verify.json the `obsidian_note` path from config.json
   ```
3. If `cells_changed` is 0: good, continue to Phase 6.
4. If non-zero: warn the user about the discrepancy and flag for manual review, but continue.

---

## Phase 6: Audit

From the extract JSON's `validation` array, flag:

1. **Uncategorized transactions** from validation notes.
2. **Large transactions** over $5,000 that may need review.
3. **Category distribution warnings** (unusual spikes or missing categories).
4. **Partial-year status** for the current year (note how many months of data exist).

---

## Phase 7: Report

Present a summary to the user:

1. **Tables updated**: count of tables changed.
2. **Cells changed**: total number of cell-level changes.
3. **Specific changes**: list each change from the diff output `changes` array (table name, field, old value, new value).
4. **Audit findings**: any flags from Phase 6.

---

## Troubleshooting

- **Pipeline script not found**: Verify the `pipeline_script` path from config.json exists and is executable.
- **No master CSVs found**: Check the `master_dir` path from config.json has `*_master.csv` files.
- **Obsidian note not found**: iCloud sync may be delayed -- check the path exists.
- **Table not found in note**: The diff script couldn't find the expected `###` heading -- check the Obsidian note structure hasn't changed.