---
name: "source-command-digest"
description: "Summarize a user's shipped work across repositories for a requested date range."
metadata:
  createdAt: "2026-09-11T18:46:00.191385+00:00"
  repoUrl: "https://skillz.supply/skills/did:plc:xbtmt2zjwlrfegqvch7fboei/skills.source-command-digest/SKILL.md"
---

# source-command-digest

Use this skill when the user asks for a digest of shipped work over a date range.
Default to the last month when no range is specified.

## Command Template

generate a comprehensive digest of what i've shipped over the last month.

## step 1: detect identity

determine who the user is by running these commands:
- `git config user.email` - primary identifier for commit authorship
- `git config user.name` - display name
- `gh api user --jq .login 2>/dev/null` - github username (skip if gh unavailable)

## step 2: detect source directories

Start with workspace roots supplied by the user or configured in the environment.
These common layouts are optional discovery hints; skip any that do not exist:

**github repos**: `~/github.com/{username}` where username matches the gh login or can be inferred from directory contents

**tangled repos**: `~/tangled.org/<owner>/<repo>` — inspect owner directories relevant to the user. Also inspect legacy `~/tangled.sh`
checkouts if present; deduplicate mirrors and worktrees by remote and commit.

**gitlab repos**: `~/gitlab.com/{username}` (if exists)

for each provider, only proceed if the base directory exists (e.g., `~/github.com`).

## step 3: find work in the requested range

for each detected source directory:
1. enumerate Git checkouts, including `.git` files used by worktrees; use
   a shallow `tree` to orient before searching.
2. inspect commit history for the requested date range and known author
   identities. Directory modification time does not establish Git activity.
3. distinguish merged/released/deployed work from local-only commits using
   the available refs and release evidence. State when shipping is unverified;
   follow relevant history beyond a truncated log when needed.

also check for work contributions in common locations:
- employer repos (other directories under ~/github.com that aren't the user's username)
- look for repos where the user has commits but doesn't own the repo

## step 4: gather project metadata

for each active repo, try to determine:
- **what it is**: read README.md first paragraph or pyproject.toml description
- **live url**: use documented URLs or deployment configuration; verify
  before claiming a URL is live. An app name alone does not prove deployment.
- **package name**: check pyproject.toml `[project] name` for installable packages

## step 5: generate digest

organize the output as markdown:

```
## shipped this month

### projects
| project | description | link |
|---------|-------------|------|
| ... | ... | ... |

### contributions
notable commits to other projects (employer, open source, etc.)

### themes
- common patterns or focus areas this month
```

include valid links wherever possible - deployed URLs, github repos, pypi packages.

format should be suitable for embedding in a personal site or sharing as a status update.