decision-log

Decision recording + result tracking skill

Security Scan

The skill's requests and instructions are coherent with its stated purpose: it records decisions to local memory files and schedules an auto-review via a cron/event mechanism — it does not ask for extra credentials or install anything.

Purpose & Capability

Name and description (decision recording + 30-day review) match the SKILL.md: it writes decision markdown files under memory/decisions/ and schedules an auto-review. No unrelated binaries, env vars, or external credentials are requested.

Instruction Scope

Instructions are focused on creating decision files and publishing a small event JSON to events/*.json and scheduling a 30-day review. They do not instruct reading arbitrary system files or pulling other environment variables. Note: the skill assumes the agent can write to memory/ and events/ and that a cron/event mechanism exists — confirm where events are published and who can read them, since recorded decisions may be sensitive.

Install Mechanism

Instruction-only skill with no install spec and no code files. Nothing is written to disk by an installer; runtime behavior depends on the agent's existing capabilities.

Credentials

The skill requests no environment variables, credentials, or config paths. Its required access (writing to memory/ and events/) is proportional to the described functionality.

Persistence & Privilege

always: false (normal). The skill intends persistent behavior via scheduled auto-review (cron) and by writing files/events, which requires the agent to retain and later read those records. This is expected for the feature but means data will persist and be revisited automatically — verify retention and access control for memory/ and events/.

Assessment

This skill appears to do what it says: record decisions to memory/decisions/ and publish small event JSON files to events/. Before installing, consider: (1) the decision contents may be sensitive — confirm where memory/ and events/ are stored and who can read them; (2) determine whether the agent's event bus actually transmits events externally (check endpoints, retention, and encryption); (3) if you don't want automatic future reads/reviews, disable or avoid the cron/auto-review functionality; and (4) test with non-sensitive sample decisions first to confirm behavior and file permissions.

Skill for recording important decisions and automatically tracking results after 30 days.

Features

• Record decision content, rationale, alternatives, expected results
• Auto-review results after 30 days (cron integration)
• File save: memory/decisions/YYYY-MM-DD-{slug}.md

Usage

Trigger keywords:

• "record decision"
• "decision log"
• "made this decision"
• "why did I do this"

Example:

Record decision: Decided to separate Instagram bot account
Rationale: Distribute main account ban risk
Alternatives: Use main account, manual operation
Expected results: Increased safety, increased management complexity

Output Format

# Decision: {title}

**Date**: YYYY-MM-DD  
**Status**: Decided / Review Pending / Results Confirmed

## Decision Content
...

## Rationale
- ...
- ...

## Alternatives Considered
1. **Alternative 1**: ...
   - Pros: ...
   - Cons: ...
2. **Alternative 2**: ...

## Expected Results
- Positive: ...
- Negative: ...

## Actual Results (Auto-update after 30 days)
_Review date: YYYY-MM-DD_

---

**Decision date**: YYYY-MM-DD | **Review date**: YYYY-MM-DD (scheduled)

Auto-review (cron)

After 30 days, automatically:

1. Compare expected vs actual results
2. Extract lessons learned
3. Generate insights for similar future decisions

Event Bus Integration

Publish event when recording decision:

• Path: events/decision-YYYY-MM-DD.json
• Format:

{
  "type": "decision-logged",
  "timestamp": "2026-02-14T12:00:00Z",
  "title": "Decision title",
  "reviewDate": "2026-03-16",
  "filePath": "memory/decisions/2026-02-14-slug.md"
}

---

decision-log | 무펭이 🐧