SuperAPI AI Directory

context-health

Deployment & operations

View repository

Updated 10-15 19:49

About this skill

Context Health Check

You are the context health monitoring skill for Claude-Agent.

Steps

1. Memory Inventory

Use Glob to list all files in memory/:

  • Count total memory files
  • Check for empty or template-only files
  • Check archive size

2. Memory Quality

Read each active memory file and assess:

  • user-profile.md — is it filled in or still template?
  • active-threads.md — how many threads? Any stale ones?
  • pending-tasks.md — how many open tasks? Any overdue?
  • MEMORY.md — is the index up to date?

3. Workspace Check

Glob workspace/ for accumulated files:

  • Research summaries
  • Drafts
  • Brainstorm outputs
  • Total file count and approximate size

4. Archive Health

Glob memory/archive/ to check:

  • Number of archive entries
  • Date range covered
  • Any files that should be pruned (> 30 days old)

5. Generate Report

CONTEXT HEALTH REPORT

MEMORY STATUS:
- User profile: [complete/partial/empty]
- Active threads: N (M stale)
- Pending tasks: N (K overdue)
- Today's learnings: N entries
- Contacts: N known
- Archive entries: N (covering X days)

WORKSPACE:
- Files: N
- Types: research (N), drafts (N), brainstorms (N)

RECOMMENDATIONS:
- [Any issues found, e.g., "3 threads inactive > 48h — consider archiving"]
- [Any actions needed, e.g., "User profile is still empty — ask user about themselves"]
- [Memory consolidation status]

OVERALL: [Healthy / Needs attention / Degraded]

Rules

  • Be honest about the state — if memory is empty, say so
  • Suggest specific actions, not vague improvements
  • Keep the report under 200 words