Skill v1.0.1
currentAutomated scan100/100+2 new
version: "1.0.1" name: cco-pack description: Build an optimal context pack for the user's task — ranked file list with offset/limit suggestions, based on git state, mentioned paths, and historical patterns license: MIT allowed-tools: [Bash, Read]
Smart Context Pack
When the user runs /cco-pack or asks "what files do I need for X", build a curated, token-budget-aware list of files to load instead of guessing or reading the whole repo.
How to use
node ${CLAUDE_PLUGIN_ROOT}/src/smart-pack.js "the task description"
Or for the current git state without a description:
node ${CLAUDE_PLUGIN_ROOT}/src/smart-pack.js
Or to feed it back into automation as JSON:
node ${CLAUDE_PLUGIN_ROOT}/src/smart-pack.js --json "..."
How it ranks files
Priority order:
- Mentioned in prompt — explicit file paths or recognisable names → relevance 100
- Modified in git working tree → 85
- Historically useful in this project (high confidence, edited often) → up to 70
- Keyword match against the task → up to 70
For each file the pack also suggests offset / limit — pointing at the structurally relevant section based on file landmarks (function/class declarations near task keywords).
Token budget awareness
Pack stops adding files once it would exceed 25% of the user's effective context budget. The budget itself respects the configured model (e.g. the full 1M window for opus-4.8 — 1M is standard on Opus, no premium tier).
Presentation
After running, present the list to the user and offer to:
- Read the top N files with the suggested offset/limit
- Save the set as a template (
/cco-templates create <name>) - Drop low-relevance items if they want a tighter pack