context
Analyze project context health and completeness with the /context command.
Before executing this command, output:
⚓ shipworthy › command:
/context— analyzing context health and completeness
Run a context health analysis for this project. Invoke the context-manager skill and check these dimensions:
Process
1. Context Infrastructure
Check which standard files exist in .shipworthy/:
| File | Purpose | Status |
|---|---|---|
architecture.md | Architecture constraints and mandatory rules | exists / missing |
regression-fence.md | Anti-pattern protection (NEVER/ALWAYS rules) | exists / missing |
config.json | Project configuration and preset | exists / missing |
INDEX.md | Memory index (auto-generated) | exists / missing |
learnings/ | At least one learning file | exists / missing |
sessions/ | At least one session file | exists / missing |
decisions/ | Architecture decision records | exists / missing |
plans/ | Implementation plans | exists / missing |
Also check if CLAUDE.md exists at the project root.
2. Regression Fence Health
- Count rules (H2 headings in
regression-fence.md) - 0 rules = no protection — recommend running
/retroafter next session - 1-15 rules = healthy
- 16-20 rules = near cap, consider pruning stale entries
- Check if any rules reference files/functions that no longer exist in the codebase
3. Session Signals
- Does
.shipworthy/.session-signalsexist? - If yes, count lines — these are unprocessed signals from hooks
- If > 0: recommend running
/retroor starting a new session (auto-retro triggers at session start)
4. Context Completeness
- Estimate total characters loaded at session start (architecture spec + fence + plans + sessions + learnings)
- Show percentage of 8000-char context budget used
- Identify gaps: "You have architecture.md but no regression fence — corrections from past sessions will repeat"
5. Actionable Suggestions
Based on analysis, suggest specific next actions. Examples:
- "12 unprocessed signals — run
/retroto convert them into learnings and fence entries" - "No regression fence yet — it will be auto-created when you run
/retroafter your next work session" - "3 learning files cover overlapping topics —
/retrowill offer to consolidate them" - "Architecture spec missing — start building with
/scaffoldor thearchitecture-awarenessskill" - "CLAUDE.md has no pointer to
.shipworthy/— consider adding:Read .shipworthy/INDEX.md for full project memory"
Output Format
Present as a concise dashboard, then suggestions:
Context Health Dashboard
========================
CLAUDE.md 82 lines ✓
.shipworthy/ 6/8 standard files ✓
Regression fence 5 rules ✓
Session signals 0 unprocessed ✓
Context budget ~5,200/8,000 chars (65%) ✓
Suggestions:
1. Add decisions/ directory for architecture decision records
2. No plans/ directory — will be created when you use writing-plans skill
After presenting the dashboard, ask:
"Want me to apply any of these suggestions?"