# Sac Pack

> Build progressive disclosure context packs (default 2-hop) for precise LLM queries over architecture knowledge.

- Skill: `spillwavesolutions/sac-pack` (Agent Skill)
- Install (CLI): `npx skillmds@latest add spillwavesolutions/sac-pack`
- Raw SKILL.md: https://api.skillmd.com/api/skills/spillwavesolutions/sac-pack/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: spillwavesolutions (https://skillmd.com/u/spillwavesolutions)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/spillwavesolutions/sac-pack

---


# SAC Pack

Walks outbound `links[]` from the seed plus inbound/backlinks (files that
point at it). Inbound uses ripgrep when `rg` is on PATH; otherwise a full
scan. Same graph either way. `--no-rg` forces the scan.

```bash
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/sac_pack.py" services/order-service.md --repo . --hops 2
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/sac_pack.py" services/order-service.md --repo . --tiny
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/sac_pack.py" services/order-service.md --repo . --tiny --summary
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/sac_pack.py" services/order-service.md --repo . --mermaid
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/sac_pack.py" services/order-service.md --repo . --no-rg --json
```

`--summary` is compact card-friendly stdout (bodies off, no mermaid), fail-closed
on the same token budget. For parent topology questions, spawn
`architecture-retriever` (`sac-retrieve`) instead of running this in the parent.