Skillshub ice-crawler-harvester
Run ICE-Crawler’s Frost→Glacier→Crystal pipeline to ingest repositories safely, emit bounded artifact bundles, and hand off sealed fossils for downstream agents.
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 "$T/skills/jacksonjp0311-gif/Clawbot-skills/ice-crawler-harvester" ~/.claude/skills/comeonoliver-skillshub-ice-crawler-harvester && rm -rf "$T"
manifest:
skills/jacksonjp0311-gif/Clawbot-skills/ice-crawler-harvester/SKILL.mdsource content
ICE-Crawler Harvester
Use this skill to run the ICE-Crawler pipeline wherever you have the project cloned. Set an environment variable such as
ICE_CRAWLER_ROOT that points to your local clone of the Ice-Crawler repo and run the commands from there. The instructions below assume PowerShell, but any shell works.
Reference: references/ice-crawler-workflow.md
Prerequisites
- Python 3.10+ with Tkinter (for the UI) and Git on PATH.
- ICE-Crawler repository checked out locally (path referenced via
).ICE_CRAWLER_ROOT - Optional:
hooks if you need partitioned follow-up tasks.agentics/
Workflow 1 — Full UI Run (interactive)
cd $env:ICE_CRAWLER_ROOT- Launch the UI:
python icecrawler.py - Paste any cloneable Git URL (browse/blob URLs are normalized automatically) and press the glowing PRESS TO SUBMIT TO ICE CRAWLER button.
- Watch the phase ladder (Frost → Glacier → Crystal → Residue) and log panels update in real time.
- When the run completes, open
to inspect the fossilized artifact bundle.state/runs/run_<timestamp>/
UI Controls
- Ctrl+B toggle left ladder; Ctrl+Shift+B toggle right logs; Ctrl+J toggle terminal.
- Drag PanedWindow sashes to resize panels.
- UI never touches git; it mirrors
written by the orchestrator.ui_events.jsonl
Workflow 2 — Headless CLI Run
cd $env:ICE_CRAWLER_ROOT $run = "state/runs/run_$(Get-Date -Format 'yyyyMMdd_HHmmss')" $temp = "state/_temp_repo" New-Item -ItemType Directory -Force -Path $run | Out-Null python -m engine.orchestrator "https://github.com/openclaw/openclaw.git" $run 80 256 $temp
Arguments:
<repo_url> <state_run_dir> <max_files> <max_kb> <temp_dir>
controls the Glacier selection cap.max_files
enforces a per-file size ceiling when copying intomax_kb
.artifact/
is purged automatically; failure to delete triggers a residue violation.temp_dir
Outputs & Follow-up
— crystallized file tree (repo-relative paths preserved).state/runs/<run>/artifact/
+artifact_manifest.json
— integrity anchors for downstream tools.artifact_hashes.json
+ai_handoff/manifest_compact.json
— sealed bundle for agent prompts.root_seal.txt
,ui_events.jsonl
— truth logs for UI or automation.run_cmds.jsonl
— teardown attestation; treat violations as failures.residue_truth.json- Extraction registry — append a row to
and drop notes underskills/ice-crawler-harvester/extractions/index.jsonl
so future skills can mine algorithms/tools (seeextractions/<repo-slug>/
).extractions/README.md
Extending / Integrating
- Call the orchestrator from scripts or scheduled jobs to keep repo fossils fresh.
- Parse
to feed other skills (e.g., code summarizers, diff analyzers).artifact_manifest.json - Hook
when you need automatic partitioning of Frost metadata or Crystal artifacts into bounded tasks.agentics/ - Adjust
/max_files
per run to dial ingest size; keep limits conservative for safety.max_kb
Troubleshooting
- Missing Tkinter on Linux →
(or distro equivalent).sudo apt install python3-tk - Git credential prompts bubble through the orchestrator; ensure SSH keys or tokens are configured.
- Residue violation (
not deleted) aborts the run; rerun after manual cleanup if needed.state/_temp_repo
Follow this skill to get deterministic repo fossils with ICE-Crawler’s provenance guarantees.