---
name: conservation-radar
description: Watches new science, policy, conservation technology and practitioner discussion for one conservation place, reads the full sources, keeps only what matters to that place's interest brief, and writes a short cited digest, a running ledger and updated place notes. Use when a reserve, park, catchment, coast or site team wants a regular research and policy watch for their place, wants to set one up or change what it watches, or asks what is new for their site since the last run.
license: MIT
---

# Conservation radar

One radar watches one place. Everything it knows lives in one **radar folder**:

```text
brief.md            what the place cares about (the interest brief)
feeds.json          sources, how often to check each, match words
state.json          last run date and items already seen
ledger.md           one row per keeper, newest first
digests/DATE.md     one digest per run
notes/*.md          durable knowledge about the place
.work/              scratch space for one run; emptied at finish
```

Every markdown file carries YAML frontmatter with a `type` field, so the folder
is readable as an Open Knowledge Format (OKF) bundle. OKF tooling is optional.

## Run

1. **Find the radar folder.** Use the folder the user names, else
   `conservation-radar/` in the working directory. With no filesystem that
   persists (a chat app), follow the chat-app section of
   [setup.md](references/setup.md). The step is done when you have read
   `brief.md`, or know there is none.
2. **Set up** when `brief.md` is missing or the user wants to change what the
   radar watches: follow [setup.md](references/setup.md). Done when the brief,
   feeds, empty ledger and state exist and the feeds have been tested.
3. **Collect** candidates since the last run with
   [collection.md](references/collection.md). Done when every due source is
   checked, searched by hand, or recorded as failed.
4. **Read and judge** every candidate against the brief with
   [reading.md](references/reading.md). Done when each candidate is a keeper
   with a full record or a drop, and every keeper is in `ledger.md`.
5. **Write the digest** with [digest.md](references/digest.md).
6. **Update notes** for durable keepers with [notes.md](references/notes.md).
7. **Finish.** Save state as the Finish section of
   [collection.md](references/collection.md) describes. Tell the user the
   digest path, the items to act on, any brief lines you filled in for them to
   confirm, and any source that failed.

The script lives at `scripts/radar.py` inside this skill's folder. Call it by
its full path from anywhere.

To run on a schedule, read [scheduling.md](references/scheduling.md).

## Standing rules

- **Open the full artifact**: the paper, the policy text, the release notes,
  the thread itself. A press release or abstract is a pointer to the artifact.
- **Mark partial access.** When only the abstract or a summary was readable,
  say so in the ledger and the digest, and leave unread fields as `not read`.
- **Keep the evidence shape.** Every keeper records its method, scope
  (sample, place, time span) and evidence shape. A model is a model, a
  consultation draft is a draft, a forum thread is testimony.
- **Discussion volume measures attention.** Many posts show that people care.
  Knowledge claims come from the evidence a post links to, or stand as dated
  testimony from the visible participants.
- **Context stays in the digest.** Only a keeper that changes what the place
  knows or must do reaches `notes/`. Everything else stays in the ledger and
  digest as context.
- **The skill's own files are read-only.** All writing happens in the radar
  folder. Downloaded pages, PDF text and other scratch files go in
  `<radar>/.work/`.

## Optional companions

The radar runs on web search, web fetch and public feeds alone. Two skills
extend it when installed:

- `last30days` by Matt Van Horn (github.com/mvanhorn/last30days-skill) for
  wider social discussion; see collection.md.
- OKF skills (github.com/scaccogatto/okf-skills) to validate or visualise the
  radar folder as a knowledge bundle.