personal memory agent
refactor: TODO Detector loads transcripts via sol call instead of sources injection
Removes the sources config from the instruction map — the only cogitate muse
that used it. The agent now reads activity transcripts through sol call
transcripts read like every other cogitate, giving it the ability to bail
early on sparse activities and keeping the cogitate contract consistent.
Also adds now:true (was the only cogitate missing it).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
+23
-11
+23
-11
apps/todos/muse/todo.md
+23
-11
apps/todos/muse/todo.md
···
11
11
"group": "Todos",
12
12
"instructions": {
13
13
"system": "journal",
14
-
"sources": {"transcripts": true, "percepts": false, "agents": {"screen": true}},
15
14
"facets": true,
15
+
"now": true,
16
16
"activity": true
17
17
}
18
18
···
27
27
1. **Detect** new todo items from the activity transcript — commitments, action items, and reminders that represent open future work
28
28
2. **Validate** existing open todos against what happened in this activity — mark items complete when you find clear evidence
29
29
30
-
Use the Activity Context and Activity State Per Segment sections above to understand what this activity involves and to focus on relevant content in the transcript.
30
+
Use the Activity Context and Activity State Per Segment sections above to understand what this activity involves and to focus your transcript reads on relevant content.
31
31
32
32
## Tooling
33
33
···
37
37
- `sol call todos done LINE_NUMBER` – mark an entry complete
38
38
- `sol call todos upcoming` – view upcoming todos to avoid duplicates
39
39
40
-
### Transcript Commands (for deeper investigation)
41
-
- `sol call transcripts read --segment SEGMENT_KEY --full` – read full transcript for a specific segment (segment keys from this activity: $activity_segments)
40
+
### Transcript Commands
41
+
- `sol call transcripts read --segment SEGMENT_KEY --transcripts` – read audio transcript for a segment
42
+
- `sol call transcripts read --segment SEGMENT_KEY --agents` – read agent outputs (screen summaries) for a segment
43
+
- `sol call transcripts read --segment SEGMENT_KEY --full` – read everything for a segment
42
44
- `sol call journal search QUERY` – cross-reference journal content
45
+
46
+
Activity segments: $activity_segments
43
47
44
48
**Query syntax**: Terms are AND'd by default; use `OR` for alternatives, quote phrases for exact matches, append `*` for prefix matching.
45
49
46
50
## Process
47
51
48
-
### Step 1: Read the Full Activity Arc First
52
+
### Step 1: Load Transcript and Current State
49
53
50
-
**CRITICAL**: Read the entire activity transcript before making any changes. You need to understand the complete arc of what happened to distinguish:
54
+
Read the activity's transcript and current todo state before making any changes.
55
+
56
+
1. For each segment in this activity ($activity_segments), load the transcript:
57
+
`sol call transcripts read --segment SEGMENT_KEY --transcripts`
58
+
Also load screen agent output for context:
59
+
`sol call transcripts read --segment SEGMENT_KEY --agents`
60
+
2. Call `sol call todos list` to see the current checklist
61
+
3. Call `sol call todos upcoming -l 50` to check what's already scheduled
62
+
63
+
If the transcript is sparse or clearly has no actionable content (e.g., silent coding, background music), skip to Output.
64
+
65
+
### Step 2: Read the Full Activity Arc
66
+
67
+
**CRITICAL**: Read through the entire transcript before making any changes. You need to understand the complete arc of what happened to distinguish:
51
68
- Items that were **mentioned and left open** → these are todos
52
69
- Items that were **mentioned and then completed** within this activity → these are NOT todos
53
70
- Items that were **already on the checklist and completed** during this activity → mark done
54
71
55
72
For example, if someone says "I need to fix the auth flow" at 10:15 and then spends 10:15–10:45 fixing it, that is NOT a todo — it was resolved within the activity.
56
-
57
-
### Step 2: Load Current State
58
-
59
-
1. Call `sol call todos list` to see the current checklist
60
-
2. Call `sol call todos upcoming -l 50` to check what's already scheduled
61
73
62
74
### Step 3: Detect New Todos
63
75