learning-capture

Name: learning-capture
Author: Svenja-dev

4installs5forks

How to install learning-capture?

svenja-dev/claude-code-skills/learning-capture

$npx -y skills add Svenja-dev/claude-code-skills --skill learning-capture

Installs into the current project.

Summary

Recognize and capture reusable patterns, workflows, and domain knowledge from work sessions into new skills. Use when completing tasks that involve novel approaches repeated 2+ times, synthesizing complex domain knowledge across conversations, discovering effective reasoning patt

Files · 1

View on GitHub

SKILL.md

1	# Learning Capture
2
3	## Overview
4
5	This skill enables continual learning by recognizing valuable patterns during work and capturing them as new skills. It focuses on high-ROI captures: patterns that will save significant context window tokens through frequent reuse.
6
7	## Recognition Framework
8
9	Monitor for these five types of learning moments:
10
11	### 1. Novel Problem-Solving Approaches
12	Trigger: Develop a creative, non-obvious solution to a complex problem that could apply to similar future problems.
13
14	Strong signals:
15	- Solution required multi-step reasoning or novel tool combinations
16	- Approach is generalizable beyond this specific instance
17	- User expresses satisfaction with the results
18	- Similar problem type likely to recur
19
20	### 2. Repeated Patterns
21	Trigger: User requests similar tasks 2-3 times and a consistent approach emerges.
22
23	Strong signals:
24	- Pattern has repeated 2+ times with consistent structure
25	- User asks "can you do the same thing as before?"
26	- Task type is clearly ongoing (e.g., weekly reports, monthly communications)
27	- Each instance requires re-explaining the approach
28
29	### 3. Domain-Specific Knowledge
30	Trigger: User explains company processes, terminology, schemas, or standards that span multiple conversations.
31
32	Strong signals:
33	- Information accumulates across 2+ conversations
34	- Knowledge is stable (won't change weekly)
35	- User frequently asks questions in this domain
36	- Re-explaining costs 1000+ tokens each time
37
38	### 4. Effective Reasoning Patterns
39	Trigger: Discover a particular way of structuring thinking that consistently produces better results.
40
41	Strong signals:
42	- Pattern applies to a category of problems, not just one instance
43	- Results are notably better than simpler approaches
44	- Structure is teachable and reproducible
45	- Problem category recurs frequently
46
47	### 5. Workflow Optimizations
48	Trigger: Figure out an efficient way to chain tools or steps together that produces comprehensive results.
49
50	Strong signals:
51	- Workflow chains 3+ distinct steps
52	- Pattern generalizes to similar task types
53	- User appreciates the thoroughness
54	- Similar workflows likely needed regularly
55
56	## Decision Framework
57
58	Offer capture when ALL of the following are true:
59
60	1. High confidence (>95%) of significant ROI:
61	- Pattern will be reused 10+ times across future conversations
62	- Each reuse saves 500+ tokens of re-explanation
63	- The skill itself costs <5000 tokens to load
64
65	2. Strong reusability signal present:
66	- Pattern has repeated 2+ times already, OR
67	- User explicitly indicates ongoing need ("I do this weekly"), OR
68	- Complex domain knowledge worth formalizing, OR
69	- Novel workflow with clear generalizability
70
71	3. Not redundant with existing capabilities:
72	- No existing skill already covers this pattern
73	- Adds meaningful value beyond general knowledge
74
75	Do NOT offer capture when:
76	- First instance of a pattern (wait for repetition)
77	- Highly context-specific solution (won't generalize)
78	- Simple task using existing capabilities (no marginal value)
79	- Creative/one-off work (low reuse probability)
80	- Ambiguous reusability (unclear if it will recur)
81
82	Consult references/decision-examples.md for concrete examples of high-confidence vs. low-confidence scenarios.
83
84	## Capture Process
85
86	### Step 1: Recognize the Learning Moment
87
88	While working, monitor for recognition triggers from the framework above. Track:
89	- Is this a repeated pattern?
90	- Does this generalize beyond this instance?
91	- Would formalizing this save significant tokens in future uses?
92
93	### Step 2: Evaluate Against Decision Framework
94
95	Before offering capture, verify:
96	- ROI calculation: (Expected_reuses × Tokens_saved) >> Skill_cost
97	- Strong reusability signal is present
98	- Not redundant with existing capabilities
99
100	If all checks pass, proceed to offer. If uncertain, do NOT offer.
101
102	### Step 3: Offer Capture Conservatively
103
104	Timing: Offer after completing the immediate task, not mid-task.
105
106	Phrasing: Be concise and specific about what would be captured and why it's valuable.
107
108	Good examples:
109	- "I notice I've structured the last three internal comms documents similarly. Would it be helpful to capture this as a skill for future communications?"
110	- "I've built up understanding of your data architecture across our conversations. Should I formalize this as a skill for more efficient future reference?"
111	- "The validation workflow I developed seems applicable to your other messy datasets. Worth capturing as a skill?"
112
113	**

People who liked this also viewed

6 picks

Type

learning-capture

4installs5forks

View on GitHub

How to install learning-capture?

svenja-dev/claude-code-skills/learning-capture

$npx -y skills add Svenja-dev/claude-code-skills --skill learning-capture

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Learning Capture
2
3	## Overview
4
5	This skill enables continual learning by recognizing valuable patterns during work and capturing them as new skills. It focuses on high-ROI captures: patterns that will save significant context window tokens through frequent reuse.
6
7	## Recognition Framework
8
9	Monitor for these five types of learning moments:
10
11	### 1. Novel Problem-Solving Approaches
12	Trigger: Develop a creative, non-obvious solution to a complex problem that could apply to similar future problems.
13
14	Strong signals:
15	- Solution required multi-step reasoning or novel tool combinations
16	- Approach is generalizable beyond this specific instance
17	- User expresses satisfaction with the results
18	- Similar problem type likely to recur
19
20	### 2. Repeated Patterns
21	Trigger: User requests similar tasks 2-3 times and a consistent approach emerges.
22
23	Strong signals:
24	- Pattern has repeated 2+ times with consistent structure
25	- User asks "can you do the same thing as before?"
26	- Task type is clearly ongoing (e.g., weekly reports, monthly communications)
27	- Each instance requires re-explaining the approach
28
29	### 3. Domain-Specific Knowledge
30	Trigger: User explains company processes, terminology, schemas, or standards that span multiple conversations.
31
32	Strong signals:
33	- Information accumulates across 2+ conversations
34	- Knowledge is stable (won't change weekly)
35	- User frequently asks questions in this domain
36	- Re-explaining costs 1000+ tokens each time
37
38	### 4. Effective Reasoning Patterns
39	Trigger: Discover a particular way of structuring thinking that consistently produces better results.
40
41	Strong signals:
42	- Pattern applies to a category of problems, not just one instance
43	- Results are notably better than simpler approaches
44	- Structure is teachable and reproducible
45	- Problem category recurs frequently
46
47	### 5. Workflow Optimizations
48	Trigger: Figure out an efficient way to chain tools or steps together that produces comprehensive results.
49
50	Strong signals:
51	- Workflow chains 3+ distinct steps
52	- Pattern generalizes to similar task types
53	- User appreciates the thoroughness
54	- Similar workflows likely needed regularly
55
56	## Decision Framework
57
58	Offer capture when ALL of the following are true:
59
60	1. High confidence (>95%) of significant ROI:
61	- Pattern will be reused 10+ times across future conversations
62	- Each reuse saves 500+ tokens of re-explanation
63	- The skill itself costs <5000 tokens to load
64
65	2. Strong reusability signal present:
66	- Pattern has repeated 2+ times already, OR
67	- User explicitly indicates ongoing need ("I do this weekly"), OR
68	- Complex domain knowledge worth formalizing, OR
69	- Novel workflow with clear generalizability
70
71	3. Not redundant with existing capabilities:
72	- No existing skill already covers this pattern
73	- Adds meaningful value beyond general knowledge
74
75	Do NOT offer capture when:
76	- First instance of a pattern (wait for repetition)
77	- Highly context-specific solution (won't generalize)
78	- Simple task using existing capabilities (no marginal value)
79	- Creative/one-off work (low reuse probability)
80	- Ambiguous reusability (unclear if it will recur)
81
82	Consult references/decision-examples.md for concrete examples of high-confidence vs. low-confidence scenarios.
83
84	## Capture Process
85
86	### Step 1: Recognize the Learning Moment
87
88	While working, monitor for recognition triggers from the framework above. Track:
89	- Is this a repeated pattern?
90	- Does this generalize beyond this instance?
91	- Would formalizing this save significant tokens in future uses?
92
93	### Step 2: Evaluate Against Decision Framework
94
95	Before offering capture, verify:
96	- ROI calculation: (Expected_reuses × Tokens_saved) >> Skill_cost
97	- Strong reusability signal is present
98	- Not redundant with existing capabilities
99
100	If all checks pass, proceed to offer. If uncertain, do NOT offer.
101
102	### Step 3: Offer Capture Conservatively
103
104	Timing: Offer after completing the immediate task, not mid-task.
105
106	Phrasing: Be concise and specific about what would be captured and why it's valuable.
107
108	Good examples:
109	- "I notice I've structured the last three internal comms documents similarly. Would it be helpful to capture this as a skill for future communications?"
110	- "I've built up understanding of your data architecture across our conversations. Should I formalize this as a skill for more efficient future reference?"
111	- "The validation workflow I developed seems applicable to your other messy datasets. Worth capturing as a skill?"
112
113	**

People who liked this also viewed

6 picks

Type