cfsa-antigravity 2.19.2 → 2.19.3

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/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "cfsa-antigravity",
3
- "version": "2.19.2",
3
+ "version": "2.19.3",
4
4
  "description": "CFSA Pipeline — Constraint-First Specification Architecture for AI agents. Production-grade from line one.",
5
5
  "scripts": {
6
6
  "changeset": "changeset",
@@ -1,6 +1,6 @@
1
1
  # Kit Sync State
2
2
 
3
3
  upstream: https://github.com/RepairYourTech/cfsa-antigravity
4
- last_synced_commit: 15665e628973ef08a1dae1c96a9567c4170f4f66
5
- last_synced_at: 2026-04-06T05:36:53Z
6
- kit_version: 2.19.2
4
+ last_synced_commit: 0db678c52d040a4c0f8944f1986f37420aeaa250
5
+ last_synced_at: 2026-04-06T05:47:35Z
6
+ kit_version: 2.19.3
@@ -93,6 +93,36 @@ This memory system complements Claude Code's built-in features:
93
93
  - **Memory**: Decision/blocker/pattern tracking is CFSA-specific, stored here
94
94
  - **Sessions**: Session logs provide richer context than Claude Code's history
95
95
 
96
+ ## Native Memory Bridge (Required)
97
+
98
+ When running in Claude Code, memory capture is dual-write:
99
+
100
+ 1. Write CFSA operational memory to `.claude/memory/*`
101
+ 2. Also sync durable collaboration memory to Claude native memory store using the native memory types
102
+
103
+ ### Mapping Rules
104
+
105
+ - `patterns.md` entries map to native **feedback** memory when they reflect user corrections/preferences
106
+ - `decisions.md` entries map to native **project** memory for project decisions; map to **user** memory when the decision is actually a user preference
107
+ - `blockers.md` entries map to native **project** memory when they affect ongoing work context
108
+ - External system pointers discovered while logging decisions/blockers map to native **reference** memory
109
+
110
+ ### Source of Truth
111
+
112
+ - `.claude/memory/*` remains the pipeline execution log
113
+ - Native memory remains the long-lived collaboration memory across sessions
114
+ - Do not replace one with the other; keep them synchronized when trigger conditions are met
115
+
116
+ ### What Not To Sync
117
+
118
+ - Trivial implementation details (file names, one-off variable choices)
119
+ - Routine entries with no future value
120
+ - Any codebase facts directly derivable from files or git history
121
+
122
+ ### Trigger
123
+
124
+ On any memory-capture trigger (decision/correction/pattern/blocker), evaluate whether a native memory entry should also be written and sync it in the same conversation.
125
+
96
126
  ## Cleanup
97
127
 
98
128
  Session logs are automatically cleaned during template build, keeping only README.md. Decision, pattern, and blocker logs persist across installations.
@@ -0,0 +1,10 @@
1
+ # Blockers
2
+
3
+ Capture active and resolved blockers that affect progress.
4
+
5
+ ## Entry format
6
+
7
+ ### BLOCKER-001: [Description] (YYYY-MM-DD)
8
+ - **Status**: active | resolved
9
+ - **Impact**: What this blocks
10
+ - **Resolution**: How it was resolved (if resolved)
@@ -0,0 +1,12 @@
1
+ # Decisions
2
+
3
+ Capture non-trivial decisions with downstream impact.
4
+
5
+ ## Entry format
6
+
7
+ ### DEC-001: [Decision summary] (YYYY-MM-DD)
8
+ - **Problem**: What needed deciding
9
+ - **Options considered**: At least 2
10
+ - **Decision**: What was chosen and why
11
+ - **Downstream**: What this affects
12
+ - **Reversibility**: High | Medium | Low
@@ -0,0 +1,12 @@
1
+ # Patterns
2
+
3
+ Capture repeatable best practices and anti-patterns discovered during execution.
4
+
5
+ ## Entry format
6
+
7
+ ### PAT-001: [Short description] (YYYY-MM-DD)
8
+ - **Type**: best-practice | anti-pattern
9
+ - **Confidence**: 0.5
10
+ - **Context**: Where this applies
11
+ - **Pattern**: What to do (or avoid)
12
+ - **Source**: What triggered this entry
@@ -0,0 +1,5 @@
1
+ # Session Logs
2
+
3
+ Store session-close notes here for continuity between runs.
4
+
5
+ Recommended file naming: `YYYY-MM-DD.md`.
@@ -37,7 +37,7 @@ A unit of work is only DONE when:
37
37
  - Phase progress fractions updated (e.g., `3/10` → `4/10`)
38
38
  - Overall progress fractions updated
39
39
  4. **The Locks**: All task claims (`[!]` flags and `files:` blocks) are removed.
40
- 5. **The Memory**: All three memory files in `.agent/progress/memory/` are checked:
40
+ 5. **The Memory**: All three memory files in `.claude/memory/` are checked:
41
41
  - `patterns.md` — Any patterns observed (what worked, what failed) logged per Protocol 04. **User corrections are logged as anti-patterns.**
42
42
  - `decisions.md` — Any non-trivial decisions logged per Protocol 06
43
43
  - `blockers.md` — Any blockers hit logged with status and impact
@@ -60,12 +60,25 @@ The pipeline has three memory files (`memory/patterns.md`, `memory/decisions.md`
60
60
  - Isolated implementation decisions (variable names, file paths) — skip
61
61
  - Repeating an existing pattern already logged — update confidence instead
62
62
 
63
+ ## Native Memory Sync (Claude Code)
64
+
65
+ When running in Claude Code, memory capture is dual-write:
66
+
67
+ 1. Write the CFSA entry to `.claude/memory/{patterns|decisions|blockers}.md`
68
+ 2. Evaluate native-memory relevance and sync to Claude native memory store:
69
+ - User correction/preference → `feedback`
70
+ - User-specific preference/profile signal → `user`
71
+ - Project decision/blocker context with downstream impact → `project`
72
+ - External system pointer/where-to-look info → `reference`
73
+
74
+ If a trigger requires CFSA logging and native-memory relevance, both writes are required in the same conversation.
75
+
63
76
  ## Pre-Completion Check
64
77
 
65
78
  Before calling `notify_user` to report completion of ANY workflow or substantial task:
66
79
 
67
80
  1. **Scan this conversation** for triggers in the table above
68
- 2. **If triggers found** → write entries to the appropriate memory files
81
+ 2. **If triggers found** → write entries to `.claude/memory/*` and sync relevant items to native memory
69
82
  3. **If no triggers** → explicitly confirm: "No new patterns, decisions, or blockers to log"
70
83
 
71
84
  This check is **not skippable**. It applies to every pipeline stage, every conversation, every session.
@@ -80,6 +80,6 @@
80
80
  ```
81
81
 
82
82
  5. **Initialize memory files** (only if they don't already exist):
83
- - `.agent/progress/memory/patterns.md` — empty with header
84
- - `.agent/progress/memory/blockers.md` — empty with header
85
- - `.agent/progress/memory/decisions.md` — empty with header
83
+ - `.claude/memory/patterns.md` — empty with header
84
+ - `.claude/memory/blockers.md` — empty with header
85
+ - `.claude/memory/decisions.md` — empty with header