| name | log-decision |
| description | Auto-triggers when a non-obvious decision is made during conversation. Also triggers on 'log decision', 'why did we choose', 'what was the trade-off', 'should we use A or B' (after resolution), or when user asks about past reasoning behind a choice. |
Log Decision -- Append-Only Decision Tracking Skill
Every important "why" captured, every trade-off remembered.
Activation
When this skill activates, silently read main/decisions.md.
- If user is logging a decision: capture and confirm
- If user is searching past decisions: find and present
- If a decision-worthy moment is detected: offer to log it
Context Guard
| Context | Status |
|---|
| User says "log decision" | ACTIVE -- capture decision |
| User says "why did we choose [X]?" | ACTIVE -- search decision log |
| Non-obvious decision made in conversation | ACTIVE -- offer to log |
| User says "should we use A or B?" (after resolved) | ACTIVE -- log the choice |
| Session end with unlogged decisions | ACTIVE -- prompt to log |
| Mid-conversation (no decision context) | DORMANT |
Protocol
On Decision Detection
On Logging a Decision
On Searching Decisions
On Reversing a Decision
Mandatory Rules
- Append-only -- NEVER edit or delete past entries
- Context + Decision + Rationale -- all three fields required for every entry
- Include rejected alternatives -- document what was NOT chosen and why
- Non-obvious only -- skip trivial choices (using git, formatting preferences)
- Natural detection -- detect decision-worthy moments without requiring explicit commands
Edge Cases
| Situation | Behavior |
|---|
| No decisions.md file | Create it with template from install protocol |
| No matching decision found | Inform user, offer to log a new entry |
| Vague decision ("we went with the better one") | Ask user to clarify alternatives and reasoning |
| Multiple related decisions | Log each as a separate entry with cross-references |
| Reversal of past decision | Append new entry with "Reversed:" prefix, never edit original |
Level History
- Lv.1 -- Base: decision detection, append-only logging, Context+Decision+Rationale format, search, reversal tracking.