competitor-watch

What it does: Once a week, fetches the public blog/changelog/release feeds of competitors you list, summarizes what changed, and writes the digest into your Obsidian vault (or a plain folder if you don’t use Obsidian). Schedule: 0 9 * * 1 — 09:00 every Monday. Risk: low-risk — fetches HTTP, writes a single new file. Never overwrites your notes. Tools used: web_search (Brave/Tavily/Exa/Perplexity/Parallel — whichever you have configured), http_request, defuddle skill (for clean text extraction), obsidian skill (optional), write_file.

Why this is useful

Knowing what your competitors shipped last week is most of the work of competitive analysis, and almost nobody actually does it weekly because reading a dozen blogs is boring. This recipe is the boring part automated. The interesting part — what it means for your roadmap — stays human.

The workflow file

---
name: competitor-watch
description: Weekly competitor changelog/blog digest, written to Obsidian.
schedule: "0 9 * * 1"
autoApprove: low-risk
catchUpOnRestart: true
maxCatchUpAge: 604800
maxIterations: 80
skills:
  - defuddle
  - digest
  - obsidian
---

# Competitor Watch — Weekly Digest

You are building a weekly digest of what my competitors shipped or said publicly in the last 7 days.

## Competitors

Edit this block to whatever you actually care about. Each entry is `name | homepage | feed/changelog URL (optional)`.

- Acme AI | https://acme.ai | https://acme.ai/changelog
- Foo Labs | https://foo.dev | https://foo.dev/blog/feed.xml
- BarCorp | https://barcorp.com |

If a feed URL is empty, fall back to a `web_search` for `site:<homepage> after:<7 days ago>`.

## Step 1 — Collect

For each competitor:

1. If a feed URL is provided, `http_request` it. Parse the items dated within the last 7 days. Keep title, URL, date.
2. If no feed URL, use `web_search` with `site:<homepage>` and a 7-day `fromDate`.
3. For each item URL, fetch the page (`http_request`) and use the `defuddle` skill to strip nav/ads/junk down to clean main content.

Cap each competitor at the 5 most recent items.

## Step 2 — Summarize

For each item, extract:

- **What** — one sentence on what was announced.
- **Why it matters** — one sentence on the implication. Be honest. "Probably nothing" is a valid answer.
- **Tag** — one of: `product`, `pricing`, `funding`, `hiring`, `partnership`, `blog`, `other`.

## Step 3 — Write the digest

Use this template, save to my Obsidian vault if `obsidian` is reachable, otherwise save to `$HOME/competitor-watch/<YYYY>/<MM>/week-of-<YYYY-MM-DD>.md` (the Monday of the week).

If using Obsidian: target path `Competitive/Weekly/Week of <YYYY-MM-DD>.md`. Use `obsidian create path=...` (do not overwrite if it exists — append a numeric suffix).

```markdown
# Competitor Watch — Week of <YYYY-MM-DD>

## Headlines
[3–5 bullets summarizing the week across competitors. Skip if quiet.]

## <Competitor name>
- **<Item title>** [<tag>]
  What: <one sentence>
  Why it matters: <one sentence>
  Source: <url> · <date>

## <Next competitor>
- ...

## Quiet this week
[List any competitors with zero items.]

## Sources
[Bullet list of every URL fetched, for traceability.]

Rules

Don’t invent items. If http_request returns 4xx/5xx for a feed, log it under a ## Errors section and skip that competitor.
Don’t editorialize beyond “Why it matters”. One sentence is enough.
Read-only on the web. Never POST anywhere.
Never overwrite an existing weekly file — suffix -2, -3, etc.


## How to install

```bash
# 1. Make sure web_search is configured. Pick one provider:
jazz config show | grep -i webSearch
# or set it through the chat with: /config

# 2. (Optional) make sure Obsidian is running and the CLI is enabled
#    (Settings → General → Command line interface)
which obsidian || echo "Obsidian CLI not found — recipe will fall back to plain files"

# 3. Drop in the workflow
mkdir -p ~/.jazz/workflows/competitor-watch
$EDITOR ~/.jazz/workflows/competitor-watch/WORKFLOW.md  # paste, then edit the competitor list

# 4. Dry run
jazz workflow run competitor-watch

# 5. Schedule
jazz workflow schedule competitor-watch

How to customize

More frequent — change schedule to 0 9 * * 1,4 (Mon and Thu).
No Obsidian — remove obsidian from the skills: list and delete the Obsidian branch in Step 3. The fallback path under $HOME/competitor-watch/ is already there.
Notion instead of Obsidian — install the upstream Notion MCP server (npx -y mcp-remote https://mcp.notion.com/mcp or makenotion/notion-mcp-server), wire it into Jazz with jazz mcp add, then ask the workflow to write the page via the Notion MCP tool. Jazz does not ship a Notion integration on its own.
Changelog-only mode — drop the homepage scraping step and require a feed URL for each competitor. Cleaner, less noise.

What you’ll see

A new note every Monday with one section per competitor and a short headline section. Over time the Competitive/Weekly/ folder becomes a searchable archive of how each competitor moved.

Limits

Costs an API key for the search provider. Brave and Tavily have free tiers. Exa, Perplexity, and Parallel are paid. Pick one and set it in your Jazz config.
Some competitor sites block scrapers. The defuddle skill handles most blogs but won’t beat a Cloudflare challenge — the recipe will skip those and surface them under ## Errors.
The Obsidian skill talks to a running Obsidian app via IPC. If the app isn’t open when the workflow fires, the fallback file path is used.