petcom/context

Context Loading Skill

Pre-implementation context loading. Combines ADR lookup, pattern loading, and memory recall.

Usage

/context                    # Auto-detect work type from conversation
/context quick              # Quick orientation (memory + recent activity only)
/context new-endpoint       # Explicit work type
/context certificates       # Search by topic

Quick Mode (/context quick)

Fast orientation for session start or context recovery:

1. Read memory/memory-log.md for recent activity
2. Read memory/context/project-overview.md
3. If topic mentioned, search memory/ for keywords
4. Output brief summary:

   ## Context Loaded
   **Recent Activity:** [from memory-log]
   **Relevant:** [matching entities/patterns]
   

Full Mode (default)

1. Detect Work Type from conversation keywords: | Work Type | Keywords | |-----------|----------| | new-endpoint | route, endpoint, api, controller | | new-model | model, schema, collection, mongoose | | bug-fix | fix, bug, issue, broken, error | | auth-change | auth, permission, access, role | | testing | test, spec, vitest, coverage | | ui-component | component, widget, page, layout |

2. Load ADRs (max 3) from dev_communication/shared/architecture/decisions/ — decision section only, skip rationale unless requested

3. Load Patterns (max 4) from memory/patterns/ matching work type

4. Load Memory — search memory/entities/ and memory/context/ for topic keywords

5. Load Backend Context Packs (backend only) — if team.json indicates backend team, load relevant notes from memory/context/backend/ (for example restart checklist or backend issue snapshots)

6. Load Role Guidance — read the role guidance file for the active sub-team from dev_communication/shared/guidance/

7. Output:

   ## Context for: {work-type}
   
   ### ADRs
   - **{ID}**: {one-line decision}
   
   ### Patterns
   - **{name}**: {summary}
   
   ### Memory
   - **{entity/context}**: {relevant notes}
   
   ### Role Guidance
   - **{key points from role guidance}**
   
   ### Checklist
   - [ ] {applicable checklist items}
   

Token Budget

Target: <2000 tokens per invocation

• Index scan: ~200 tokens
• ADR summaries: ~150 each (450 max)
• Pattern loads: ~300 each (1200 max)