Skip to content
aspens

aspens doc impact

Show whether your generated context is still keeping up with the codebase. This is the proof surface: are the docs present, covering the right domains, surfacing the right hub files, and fresher than the repo changes?

Usage

Section titled “Usage”
Terminal window
npx aspens doc impact [path]

What it checks

Section titled “What it checks”
  • Instructions present — does CLAUDE.md or AGENTS.md exist per target?
  • Skills present — how many domain skills are installed?
  • Domain coverage — which detected repo domains have matching skills?
  • Hub file surfacing — are the top hub files mentioned in root guidance?
  • Drift — are generated docs older than the newest source changes?
  • Hook health — are required hooks installed and pointing to valid scripts?
  • Save-tokens health — is the save-tokens installation complete? (Claude only)
  • Opportunities — optional aspens features not yet installed

Example output

Section titled “Example output”
$ aspens doc impact


  my-app


  Claude Code
    Instructions: present (CLAUDE.md)
    Skills: 9
    Domain coverage: 8/9, missing onboarding
    Hub files surfaced: 4/5
    Hooks: installed
    Save tokens: healthy
    Last updated: Apr 7, 2026, 9:41 AM  stale vs source

Health score

Section titled “Health score”

The health score starts at 100 and deducts for issues:

IssueDeduction
Missing instructions file-35
No skills installed-25
Domain gapsup to -25
Missed hub files-4 each
Drift (stale files)-3 per file, max -20
Unhealthy hooks-10
Broken save-tokens-5

LLM interpretation

Section titled “LLM interpretation”

If a backend CLI is available, impact can send the report to an LLM for interpretation:

Terminal window
npx aspens doc impact --backend claude

The LLM returns a structured analysis with:

  • Bottom line — one-sentence summary
  • Improvements — what aspens is doing well
  • Risks — what needs attention
  • Next step — recommended action

Interactive apply

Section titled “Interactive apply”

When the report identifies issues, aspens can fix them:

Terminal window
npx aspens doc impact
# Shows report, then prompts:
# ◆ Apply recommended repairs? (y/n)

This builds a repair plan and applies it — regenerating missing skills, reinstalling broken hooks, etc.

Options

Section titled “Options”
OptionDescription
--backend <backend>Backend for LLM interpretation
--model <model>Model for interpretation
--no-graphSkip import graph analysis
--timeout <seconds>Backend timeout
--verboseShow diagnostic output