moonlight-lupin/agent-skills / productivity/decision-log
decision-log
VerifiedUse when an architecture/product decision should be recorded or past decisions need re-checking — an ADR-style decision journal for agents and teams. Create numbered decision records, track superseding chains, schedule periodic reviews, and search past decisions to avoid re-litigating settled questions.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.9 KB |
| references/adr-format.md | markdown | 5.9 KB |
| references/review-cadence.md | markdown | 4.3 KB |
| scripts/decision_log.pyexecutable | script | 19.2 KB |
| scripts/review_checker.pyexecutable | script | 4.1 KB |
| templates/decision-template.md | markdown | 502 B |
| tests/test_decision_log.pyexecutable | script | 6.1 KB |
references/review-cadence.md
View on GitHub# Review Cadence Rules
A decision record is useful only if stale decisions are revisited. The review
section gives each active decision either a date-based review schedule or a
concrete trigger condition.
## Contents
- Cadence options
- Calendar-month calculation
- How due review detection works
- Choosing the right cadence
- Review meeting prompt
## Cadence options
### monthly
Use `monthly` for decisions that can become wrong quickly:
- experiments and pilots;
- temporary process choices;
- vendor trials;
- operational workarounds;
- decisions tied to fast-changing usage, cost, or reliability data.
Computation: decision date + 1 calendar month.
Example:
```markdown
## Review
- Cadence: monthly
- Next review: 2026-08-06
- Trigger: review earlier if weekly failure rate exceeds 2%.
```
### quarterly
Use `quarterly` as the default for active architecture, product, workflow, and
operations decisions. It is frequent enough to catch drift without making review
noise too high.
Computation: decision date + 3 calendar months.
Example:
```markdown
## Review
- Cadence: quarterly
- Next review: 2026-10-06
- Trigger: review earlier if a second team depends on this interface.
```
### annually
Use `annually` for stable decisions where the environment changes slowly:
- durable standards;
- mature dependencies;
- naming conventions;
- stable governance or documentation practices;
- low-risk decisions whose cost of change is low.
Computation: decision date + 1 calendar year.
Example:
```markdown
## Review
- Cadence: annually
- Next review: 2027-07-06
- Trigger: review earlier if the underlying dependency enters maintenance mode.
```
### on-trigger
Use `on-trigger` when a calendar date is less meaningful than an event. Leave
`Next review` blank and make the trigger concrete.
Good triggers:
- `review if monthly volume exceeds 1M records`;
- `review if write concurrency becomes a product requirement`;
- `review if the vendor changes pricing materially`;
- `review if a second deployment environment is added`;
- `review if the dependency has no security update for 12 months`.
Weak triggers:
- `review later`;
- `review if needed`;
- `review if things change`.
Example:
```markdown
## Review
- Cadence: on-trigger
- Next review:
- Trigger: review if any customer requires SSO enforcement.
```
## Calendar-month calculation
The CLI computes next review dates by adding calendar months, not fixed day
counts:
- `monthly` = +1 calendar month;
- `quarterly` = +3 calendar months;
- `annually` = +12 calendar months.
When the target month has fewer days, the day clamps to the last day of that
month:
| Start date | Cadence | Next review |
| --- | --- | --- |
| 2026-01-31 | monthly | 2026-02-28 |
| 2024-01-31 | monthly | 2024-02-29 |
| 2026-11-30 | quarterly | 2027-02-28 |
| 2026-02-28 | annually | 2027-02-28 |
This matches how people usually interpret "one month later" and avoids drifting
reviews by using fixed 30- or 90-day offsets.
## How due review detection works
`decision_log.py due-review` scans all ADR files matching
`ADR-NNN-short-title.md`, extracts `Next review: YYYY-MM-DD`, and prints records
whose next review date is today or earlier. It ignores superseded and deprecated
records.
`review_checker.py` is designed for cron schedulers. It only reports accepted
records due for review and stays silent when nothing is due. This makes it easy
to wire into a weekly reminder without producing empty messages.
## Choosing the right cadence
Use the highest cadence that still protects the decision from going stale:
1. Is the decision temporary, experimental, or high-change? Use `monthly`.
2. Is it an active normal decision with meaningful future risk? Use `quarterly`.
3. Is it stable and low-change? Use `annually`.
4. Is a measurable event more important than a date? Use `on-trigger`.
If unsure, start with `quarterly`. You can change the cadence at the first
review.
## Review meeting prompt
When an ADR is due, answer these questions:
1. Is the original context still true?
2. Did any assumption fail?
3. Did any negative consequence become unacceptable?
4. Are the rejected options now better because constraints changed?
5. Should the decision remain accepted, be superseded by a new ADR, or be
deprecated?
If the decision changes, create a new ADR and supersede the old one rather than
editing the old record into the new truth.Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |