Transaction History Builder

Transaction History Builder

Aggregates a developer's shipped work into a single, continuously updated "transaction history" -- a living portfolio that replaces the static resume with verifiable evidence of what was built, when, and why. Each entry is a transaction: a concrete artifact with context, not a credential claim.

Prerequisites

• GitHub CLI (gh) authenticated
• Access to local project directories
• Optional: Obsidian vault for explanation artifacts
• Optional: Project manifests (package.json, pyproject.toml, CLAUDE.md) for metadata

Phase 1: Source Discovery

Identify all sources of shipped work:

GitHub Repositories

# List all repos the user has pushed to in the last N months
gh repo list --limit 100 --json name,pushedAt,description,url | \
  python3 -c "import json,sys; [print(f'{r[\"name\"]} | {r[\"pushedAt\"][:10]} | {r[\"description\"] or \"(no description)\"}') for r in sorted(json.loads(sys.stdin.read()), key=lambda x: x['pushedAt'], reverse=True)]"

Local Project Directories

Scan common project locations:

• ~/projects/
• Any paths the user specifies

For each directory with a .git folder, extract:

• Repo name
• Last commit date
• Commit count by the user
• Languages (from file extensions)

Explanation Artifacts

Scan for existing comprehension artifacts:

• **/EXPLANATION.md
• **/COMPREHENSION-INTERVIEW.md
• **/CONTEXT.md
• **/docs/context/*.context.md

Ask the user: "Are there other locations where your work lives? (Deployed URLs, other machines, private repos)"

Phase 2: Transaction Extraction

For each discovered source, extract a transaction record:

### Transaction: <project_name>

**Type**: <new-project | feature | fix | refactor | tool | skill | deployment>
**Date range**: <first_commit> to <last_commit>
**Status**: <active | shipped | archived | abandoned>
**Repo**: <url or local path>

**What was built**: <one paragraph from README, CLAUDE.md, or commit messages>

**Evidence**:
- Commits: <count> over <duration>
- PRs merged: <count>
- Languages: <list>
- Tests: <yes/no, count if yes>
- Deployed: <url if applicable>

**Comprehension layer**:
- Explanation artifact: <path if exists, "MISSING" if not>
- Decision log: <path if exists>
- Architecture docs: <path if exists>

**Key contributions** (from git log):
- <most significant commit message>
- <second most significant>
- <third most significant>

Extracting "What was built"

Priority order for description:

1. README.md first paragraph
2. CLAUDE.md project description
3. package.json/pyproject.toml description field
4. Most common commit message prefix (feat:, fix:, etc.) to infer purpose

Filtering Noise

Skip transactions that are:

• Forks with zero original commits
• Repos with only an initial commit and no follow-up
• Config-only repos (dotfiles, etc.) unless the user explicitly wants them
• Repos last touched >12 months ago (archive them in a separate section)

Phase 3: Trajectory Analysis

After extracting all transactions, produce trajectory insights:

## Trajectory Summary

**Active period**: <earliest transaction> to <latest>
**Total transactions**: <count>
**Active projects**: <count currently active>

### Technology trajectory
- <language/framework>: <count> projects, <trend: growing/stable/declining>
- ...

### Complexity trajectory
- Average project duration: <N weeks/months>
- Longest sustained project: <name, duration>
- Multi-contributor projects: <count>

### Comprehension coverage
- Projects with explanation artifacts: <X of Y> (<percentage>%)
- Projects with test suites: <X of Y>
- Projects with architecture docs: <X of Y>
- **Comprehension gap**: <list of active projects missing explanation artifacts>

### Shipping cadence
- Transactions per month (last 6 months): <chart or numbers>
- Average time from first commit to "shipped" status: <duration>

Phase 4: Output Formats

Ask the user which format(s) they want:

Option A: Markdown Portfolio

Write to ~/vault/career/transaction-history.md (or user-specified path). Single file, full detail, suitable for Obsidian or GitHub.

Option B: JSON Index

Write to ~/.transaction-history.json. Machine-readable, suitable for static site generators or dashboards.

{
  "generated": "2026-04-20",
  "transactions": [
    {
      "name": "project-name",
      "type": "new-project",
      "date_range": ["2026-01-15", "2026-03-22"],
      "status": "shipped",
      "repo": "https://github.com/user/repo",
      "languages": ["Python", "TypeScript"],
      "commits": 142,
      "has_explanation_artifact": true,
      "has_tests": true,
      "description": "..."
    }
  ],
  "trajectory": { ... }
}

Option C: GitHub Pages Portfolio

Create a minimal static site using the JSON index:

• Create repo <user>/portfolio (or specified name)
• Generate index.html with transaction cards, trajectory charts
• Deploy to GitHub Pages
• Return the public URL

Phase 5: Incremental Updates

The transaction history should be updatable, not rebuilt from scratch each time.

On subsequent runs:

1. Read the existing history file
2. Check for new commits, new repos, new artifacts since the last run
3. Append new transactions and update existing ones
4. Re-calculate trajectory analysis
5. Note what changed: "Added 3 new transactions, updated 2 existing"

Store a last_updated timestamp in the output file.

Verification

A good transaction history has:

• Every active project represented (no gaps from "I forgot about that repo")
• Each transaction has a concrete description (not "various improvements")
• Comprehension coverage calculated and gaps flagged
• Abandoned projects honestly marked as abandoned (not hidden)
• Trajectory analysis based on actual data, not narrative ("I'm becoming a backend developer" must be backed by transaction counts)

Source Attribution

Extracted from Nate Kadlac's newsletter digest (2026-04-20) -- "Transactions over Credentials" principle. A transaction history replaces static resumes with verifiable evidence of shipped work. Nate's TalentBoard is the SaaS version; this skill builds a self-hosted, developer-controlled alternative. Idea #4: Transaction History Builder (Portfolio Aggregator).