@danielmarbach/mnemonic-mcp 0.27.0 → 0.27.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md
CHANGED
|
@@ -6,6 +6,12 @@ The format is loosely based on Keep a Changelog and uses semver-style version he
|
|
|
6
6
|
|
|
7
7
|
## [Unreleased]
|
|
8
8
|
|
|
9
|
+
## [0.27.1] - 2026-04-28
|
|
10
|
+
|
|
11
|
+
### Changed
|
|
12
|
+
|
|
13
|
+
- `mnemonic-rpi-workflow` skill: Review phase now retrieves research and plan context before evaluating implementation, with comparison guardrails against original requirements and deliverables. Plan phase includes a self-check after drafting to catch placeholders and coverage gaps. Review evidence must be gathered fresh rather than reused from implementation. Added review handoff variant for subagent-driven review with full artifact chain.
|
|
14
|
+
|
|
9
15
|
## [0.27.0] - 2026-04-28
|
|
10
16
|
|
|
11
17
|
### Added
|
package/package.json
CHANGED
|
@@ -40,6 +40,7 @@ Only after confirmation: proceed to Plan checklist.
|
|
|
40
40
|
- For non-trivial work, include a short markdown checkbox list (`- [ ]`) for executable steps.
|
|
41
41
|
- One current plan per request; update or supersede as needed.
|
|
42
42
|
- Update plan note before continuing if scope, architecture, dependencies, or assumptions change materially.
|
|
43
|
+
- After drafting, run a self-check: does each research requirement map to a plan item? Are there placeholders (TBD, TODO)? Are step references internally consistent?
|
|
43
44
|
|
|
44
45
|
### 2a. Plan → Implement Handoff
|
|
45
46
|
|
|
@@ -59,12 +60,30 @@ Only after confirmation: proceed to Implement checklist.
|
|
|
59
60
|
|
|
60
61
|
### 4. Review
|
|
61
62
|
|
|
62
|
-
|
|
63
|
-
|
|
64
|
-
|
|
65
|
-
|
|
66
|
-
|
|
67
|
-
-
|
|
63
|
+
Before reviewing, retrieve the context that informed implementation:
|
|
64
|
+
1. `get` the research note(s) and plan note linked to this work
|
|
65
|
+
2. Read the apply/task note(s) to confirm what actually shipped
|
|
66
|
+
|
|
67
|
+
Create review notes: `role: review`, `lifecycle: temporary`.
|
|
68
|
+
Link to apply/task notes or plan (`derives-from` when conclusions derive from specific artifacts).
|
|
69
|
+
|
|
70
|
+
Review against research and plan:
|
|
71
|
+
- Does the implementation satisfy the requirements identified in research?
|
|
72
|
+
- Were all planned deliverables completed? If not, why?
|
|
73
|
+
- Are there gaps between what was planned and what was delivered?
|
|
74
|
+
- Were any assumptions from research invalidated during implementation?
|
|
75
|
+
|
|
76
|
+
Before recording outcome, run a self-review checklist:
|
|
77
|
+
- Re-read research requirements — is each addressed in implementation or explicitly deferred?
|
|
78
|
+
- Re-read plan deliverables — do all checkboxes have matching verification evidence?
|
|
79
|
+
- Is any unchecked item silently ignored rather than called out?
|
|
80
|
+
|
|
81
|
+
For non-trivial work, dispatch a subagent with the full artifact chain (research, plan, apply notes) using the [review handoff variant](#review-handoff-variant) below.
|
|
82
|
+
|
|
83
|
+
Record outcome: continue, block, or update plan.
|
|
84
|
+
Reconcile checklist state with verification evidence; call out any unchecked items explicitly.
|
|
85
|
+
If review causes a material plan change, update plan note first.
|
|
86
|
+
Every verification command must be run fresh during this review — do not reuse results from implementation. Every review note must include verification evidence:
|
|
68
87
|
|
|
69
88
|
```text
|
|
70
89
|
- Command: <run command>
|
|
@@ -96,6 +115,25 @@ Task scope:
|
|
|
96
115
|
Must return: apply note content, optional review note, recommendation (continue | block | update plan)
|
|
97
116
|
```
|
|
98
117
|
|
|
118
|
+
### Review Handoff Variant
|
|
119
|
+
|
|
120
|
+
For subagent-driven review, include the full artifact chain so the reviewer has the same context as the implementer:
|
|
121
|
+
|
|
122
|
+
```text
|
|
123
|
+
Request note: <note-id/title>
|
|
124
|
+
Research notes: <note-id/title>, ...
|
|
125
|
+
Plan note: <note-id/title>
|
|
126
|
+
Apply/task notes: <note-id/title>, ...
|
|
127
|
+
Durable context: <note-id/title>, ...
|
|
128
|
+
|
|
129
|
+
Review scope:
|
|
130
|
+
- What was planned: <summary from plan note>
|
|
131
|
+
- What was implemented: <summary from apply note>
|
|
132
|
+
- Validation: <tests/checks>
|
|
133
|
+
|
|
134
|
+
Instructions: Compare implementation against research requirements and plan deliverables. Identify gaps, regressions, or deviations. Return: review note content, recommendation (continue | block | update plan), and any unchecked items.
|
|
135
|
+
```
|
|
136
|
+
|
|
99
137
|
## Canonical Graph
|
|
100
138
|
|
|
101
139
|
```text
|