gm-cc 2.0.649 → 2.0.650
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/.claude-plugin/marketplace.json +1 -1
- package/agents/memorize.md +14 -57
- package/package.json +1 -1
- package/plugin.json +1 -1
|
@@ -4,7 +4,7 @@
|
|
|
4
4
|
"name": "AnEntrypoint"
|
|
5
5
|
},
|
|
6
6
|
"description": "State machine agent with hooks, skills, and automated git enforcement",
|
|
7
|
-
"version": "2.0.
|
|
7
|
+
"version": "2.0.650",
|
|
8
8
|
"metadata": {
|
|
9
9
|
"description": "State machine agent with hooks, skills, and automated git enforcement"
|
|
10
10
|
},
|
package/agents/memorize.md
CHANGED
|
@@ -1,39 +1,18 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: memorize
|
|
3
|
-
description: Background memory agent. Classifies context
|
|
3
|
+
description: Background memory agent. Classifies context and writes to AGENTS.md + rs-learn. No memory dir, no MEMORY.md.
|
|
4
4
|
agent: true
|
|
5
5
|
---
|
|
6
6
|
|
|
7
7
|
# Memorize — Background Memory Agent
|
|
8
8
|
|
|
9
|
-
|
|
9
|
+
Writes facts to two places only: **AGENTS.md** (non-obvious technical caveats) and **rs-learn** (all classified facts via fast ingest).
|
|
10
10
|
|
|
11
|
-
|
|
12
|
-
- **Memory dir** = `<HOME>/.claude/projects/<slug>/memory/` where `<slug>` is `process.cwd()` with `:` stripped and `/` or `\\` replaced by `--` (example: `C:/dev/devbox/spawnpoint` → `C--dev-devbox-spawnpoint`). On Windows `<HOME>` = `C:/Users/<user>`; on POSIX it is `$HOME`.
|
|
11
|
+
Resolve at start of every run:
|
|
13
12
|
|
|
14
|
-
|
|
13
|
+
- **Project root** = `process.cwd()` when invoked. `AGENTS.md` is `<project root>/AGENTS.md`.
|
|
15
14
|
|
|
16
|
-
## STEP 1:
|
|
17
|
-
|
|
18
|
-
Read memory dir contents and MEMORY.md index. Read all existing memory files. Read AGENTS.md in full.
|
|
19
|
-
|
|
20
|
-
If memory dir does not exist, create it. If MEMORY.md does not exist, create it empty.
|
|
21
|
-
|
|
22
|
-
## STEP 2: PRUNE
|
|
23
|
-
|
|
24
|
-
Run BEFORE writing any new content.
|
|
25
|
-
|
|
26
|
-
Remove entries that are:
|
|
27
|
-
- Contradicted or superseded by current AGENTS.md or observable codebase facts
|
|
28
|
-
- Duplicates of other entries (merge into one)
|
|
29
|
-
- Derivable at runtime via exec:codesearch: file paths, function names, API shapes, architecture patterns
|
|
30
|
-
- Active task state, current progress, session narration
|
|
31
|
-
|
|
32
|
-
Keep ONLY: cross-session unknowns that required multiple failed attempts to discover, user preferences and feedback, project decisions with non-obvious rationale.
|
|
33
|
-
|
|
34
|
-
When in doubt: DELETE. Memory must stay lean.
|
|
35
|
-
|
|
36
|
-
## STEP 3: CLASSIFY
|
|
15
|
+
## STEP 1: CLASSIFY
|
|
37
16
|
|
|
38
17
|
Examine the ## CONTEXT TO MEMORIZE section at the end of this prompt. For each fact, classify as:
|
|
39
18
|
|
|
@@ -43,50 +22,28 @@ Examine the ## CONTEXT TO MEMORIZE section at the end of this prompt. For each f
|
|
|
43
22
|
- reference: pointers to external systems, URLs, paths
|
|
44
23
|
|
|
45
24
|
Discard:
|
|
46
|
-
- Facts already covered in AGENTS.md (exact or paraphrase)
|
|
47
25
|
- Obvious facts derivable from reading the code
|
|
48
26
|
- Active task state or session progress
|
|
27
|
+
- Facts that would not be useful in a future session
|
|
49
28
|
|
|
50
|
-
## STEP
|
|
29
|
+
## STEP 2: INGEST INTO RS-LEARN
|
|
51
30
|
|
|
52
|
-
For each classified fact:
|
|
53
|
-
1. Check all existing memory files — if one covers the same topic, merge the new fact in
|
|
54
|
-
2. If no existing file covers it, create memory/<slug>.md
|
|
31
|
+
For each classified fact, run (one call per fact):
|
|
55
32
|
|
|
56
|
-
|
|
33
|
+
```bash
|
|
34
|
+
bun x rs-learn add "<fact text>" --source "<type>/<slug>" --no-extract
|
|
57
35
|
```
|
|
58
|
-
---
|
|
59
|
-
name: <descriptive name>
|
|
60
|
-
description: <one-line under 80 chars>
|
|
61
|
-
type: user|feedback|project|reference
|
|
62
|
-
---
|
|
63
|
-
|
|
64
|
-
<body>
|
|
65
|
-
```
|
|
66
|
-
|
|
67
|
-
For feedback and project types, body must include:
|
|
68
|
-
- The fact or rule
|
|
69
|
-
- Why: <reason>
|
|
70
|
-
- How to apply: <concrete application>
|
|
71
|
-
|
|
72
|
-
## STEP 5: UPDATE MEMORY.md
|
|
73
|
-
|
|
74
|
-
Rewrite the MEMORY.md index to reflect all current files in the memory dir.
|
|
75
|
-
|
|
76
|
-
Format: one line per file, under 150 chars each:
|
|
77
|
-
`- [Title](file.md) — one-line hook`
|
|
78
|
-
|
|
79
|
-
No frontmatter. Max 200 lines.
|
|
80
36
|
|
|
81
|
-
|
|
37
|
+
Where `<fact text>` is a self-contained one-to-three sentence summary of the fact, and `<slug>` is a short kebab-case label (e.g. `feedback/terse-responses`, `project/merge-freeze`).
|
|
82
38
|
|
|
83
|
-
|
|
39
|
+
Use `--no-extract` always — fast path, no LLM overhead, ~1s per call.
|
|
84
40
|
|
|
85
|
-
## STEP
|
|
41
|
+
## STEP 3: AGENTS.md
|
|
86
42
|
|
|
87
43
|
A non-obvious technical caveat qualifies if it required multiple failed runs to discover and would not be apparent from reading code or docs.
|
|
88
44
|
|
|
89
45
|
For each qualifying fact from context:
|
|
46
|
+
- Read AGENTS.md first if not already read this run
|
|
90
47
|
- If AGENTS.md already covers it → skip
|
|
91
48
|
- If genuinely non-obvious → append to the appropriate section
|
|
92
49
|
|
package/package.json
CHANGED