Skillshub harvest-proposal-pipeline

Automate ICE-Crawler ingestion → registry update → proposal stub creation. Use when you want a full harvest loop that ends with a ready-to-review skill idea.

install

source · Clone the upstream repo

git clone https://github.com/ComeOnOliver/skillshub

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/ComeOnOliver/skillshub "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/jacksonjp0311-gif/Clawbot-skills/harvest-proposal-pipeline" ~/.claude/skills/comeonoliver-skillshub-harvest-proposal-pipeline && rm -rf "$T"

manifest: skills/jacksonjp0311-gif/Clawbot-skills/harvest-proposal-pipeline/SKILL.md

source content

Harvest → Proposal Pipeline

This skill runs the whole flow when you say so:

ICE-Crawler orchestrator ingests a repo (Frost→Glacier→Crystal→Residue).
The extraction registry (
```
ice-crawler-harvester/extractions
```
) is appended automatically.
A summary stub + optional proposal JSON is generated so you can approve the next skill.

Helper script:

scripts/harvest_pipeline.py

Usage

cd <skills_repo>\skills\harvest-proposal-pipeline
python scripts\harvest_pipeline.py <repo_url> --candidate <skill-name> --max-files 80 --max-kb 256

```
repo_url
```
— cloneable Git URL (raw/tree/blob URLs are fine; orchestrator normalizes).
```
--candidate
```
(optional) — proposed skill name. When supplied, a stub JSON is written under
```
skills/extraction-proposer/proposals/
```
.
```
--max-files
```
/
```
--max-kb
```
— bounds for Glacier selection + per-file size.
Set
```
ICE_CRAWLER_ROOT
```
to the local path of your ICE-Crawler clone before running this script.

What it does

Runs

python -m engine.orchestrator …

inside

$env:ICE_CRAWLER_ROOT

with timestamped

state/runs/run_<ts>

Loads
```
artifact_manifest.json
```
to count files.
Writes/updates
```
skills/ice-crawler-harvester/extractions/index.jsonl
```
with: repo, run_dir, manifest path, file count, summary path.
Creates/updates
```
skills/ice-crawler-harvester/extractions/<repo-slug>/SUMMARY.md
```
, including an auto-generated “Auto-detected Candidates” section with suggested algorithms/tools.
If
```
--candidate
```
is provided, saves
```
skills/extraction-proposer/proposals/<skill-name>.json
```
prefilled with provenance, candidate description, notable file list, and suggested skill structure.

After the run

Console output highlights:
- ```
Run folder
```
  : the ICE-Crawler fossil (
```
state/runs/run_<timestamp>
```
  ).
- ```
Summary
```
  : path to
```
SUMMARY.md
```
  (includes auto-detected candidate bullets).
- ```
Proposal stub
```
  : JSON file to review in
```
extraction-proposer/proposals/
```
  (if
```
--candidate
```
  supplied).
Once you approve a proposal, copy it into
```
extraction-proposer/catalog/
```
and mark the status (
```
approved
```
,
```
rejected
```
, etc.) so the catalog becomes the definitive list of green-lit algorithms.
Open the summary/proposal, add any extra notes, and decide whether to execute the build using
```
skill-creator
```
.

Safety & Notes

Requires Python + git on PATH (same prerequisites as ICE-Crawler).
Registry + proposal folders must exist (created by earlier skills).
Script fails fast if orchestrator errors or the manifest is missing.
You stay in control: nothing becomes a skill until you approve/edit the generated proposal.

Invoke this skill whenever you want a one-command harvest that comes back with a ready-to-review idea.