self-improvement-loops

Name: self-improvement-loops
Author: muratcankoylan

20installs1.4kforks

How to install self-improvement-loops?

muratcankoylan/agent-skills-for-context-engineering/self-improvement-loops

$npx -y skills add muratcankoylan/Agent-Skills-for-Context-Engineering --skill self-improvement-loops

Installs into the current project.

Summary

This skill should be used when the harness, scaffold, workflow, or optimizer itself is the optimization target: recursive self-improvement (RSI) loops, meta-harnesses, self-improving harnesses that mine their own failures and propose bounded edits, evolutionary or population-base

Files · 1

View on GitHub

SKILL.md

1	# Self-Improvement Loops
2
3	This skill covers systems where the harness is the artifact being optimized: an agent mines its own failures and edits its own scaffold, a meta-agent searches over harness code, a population of workflow candidates evolves against an evaluator, or the mechanism that produces context is itself versioned and improved. The design question shifts from "how do I control one loop" (harness-engineering) to "how do I let a loop rewrite parts of itself without corrupting the signal that steers it".
4
5	The controlling constraint across every published system: the loop optimizes whatever signal it is given, including the signal's own weaknesses. Design the loop assuming the optimizer will find every gap between the metric and the intent.
6
7	## When to Activate
8
9	Activate this skill when:
10
11	- Building a loop where an agent proposes edits to its own harness, prompts, context playbook, or workflow based on mined failure patterns
12	- Designing meta-level search over harness or scaffold code: meta-agent search, tree search over workflow graphs, evolutionary program search with an LLM mutation operator
13	- Choosing acceptance criteria for any self-modifying agent system
14	- Evolving the mechanism that manages context (a skill, playbook, or context function) rather than hand-editing the context artifact
15	- Diagnosing a degenerating self-improvement loop: reward hacking, diversity collapse, context collapse, or silent stagnation
16	- Deciding which level of the optimization ladder (prompt, context, workflow, harness code, optimizer code) a recurring failure should be fixed at
17
18	Do not activate this skill for adjacent work owned by other skills:
19
20	- Governance of a single autonomous loop that does not modify itself: locked and editable surfaces, durable logs, rollback, novelty gates, PR preparation, and human approval boundaries belong to `harness-engineering`. That skill defines the control surfaces; this skill defines what happens when the surfaces themselves become the optimization target.
21	- Building the evaluator, regression suite, or quality gates that score candidates: `evaluation`.
22	- LLM-as-judge design, pairwise comparison, and bias mitigation for candidate scoring: `advanced-evaluation`.
23	- One-shot token efficiency, masking, or caching without an improvement loop: `context-optimization`.
24	- Remote sandboxes and background execution infrastructure for running the loop: `hosted-agents`.
25	- Whether to build the loop at all, pipeline shape, and cost estimation: `project-development`.
26
27	## Core Concepts
28
29	### The Optimization Ladder
30
31	Published self-improvement systems target progressively deeper objects:
32
33	\| Rung \| Optimized object \| Example systems \|
34	\| --- \| --- \| --- \|
35	\| 1 \| Instruction prompts \| Promptbreeder, GEPA \|
36	\| 2 \| Structured context \| ACE (playbook of itemized bullets) \|
37	\| 3 \| Context mechanism \| MCE (the skill that produces context) \|
38	\| 4 \| Workflow graph \| ADAS, AFlow \|
39	\| 5 \| Harness code \| Self-Harness (bounded self-edits), Meta-Harness, Darwin Godel Machine \|
40	\| 6 \| Optimizer code \| STOP (the improver improves the improver) \|
41
42	Each rung up buys a larger design space and more leverage, and costs more per evaluation with a larger surface for gaming. Fix a recurring failure at the lowest rung that can express the fix. A stale-library failure is a context fix, not a workflow rewrite. Only move up when failure clusters at the current rung persist across candidates.
43
44	### The Loop Is Not the Intelligence
45
46	Recursive structure amplifies base-model quality in both directions. The same recursive improver design that compounds gains with a strong model degrades mean performance across iterations with weaker models (claim-self-improvement-capability-threshold). Two implications:
47
48	1. Run a capability validation before enabling recursion: a fixed number of iterations on a held-out task set, with the loop counted as net-negative if the trajectory declines.
49	2. Detect the degenerate stagnation mode. A broken improver that silently returns its input unchanged looks like stability in aggregate metrics. Track edit diffs, not only scores.
50
51	### The Outside-the-Loop Invariant
52
53	The evaluator, its instrumentation, permission control, and budget enforcem

People who liked this also viewed

6 picks

Type

self-improvement-loops

20installs1.4kforks

View on GitHub

How to install self-improvement-loops?

muratcankoylan/agent-skills-for-context-engineering/self-improvement-loops

$npx -y skills add muratcankoylan/Agent-Skills-for-Context-Engineering --skill self-improvement-loops

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Self-Improvement Loops
2
3	This skill covers systems where the harness is the artifact being optimized: an agent mines its own failures and edits its own scaffold, a meta-agent searches over harness code, a population of workflow candidates evolves against an evaluator, or the mechanism that produces context is itself versioned and improved. The design question shifts from "how do I control one loop" (harness-engineering) to "how do I let a loop rewrite parts of itself without corrupting the signal that steers it".
4
5	The controlling constraint across every published system: the loop optimizes whatever signal it is given, including the signal's own weaknesses. Design the loop assuming the optimizer will find every gap between the metric and the intent.
6
7	## When to Activate
8
9	Activate this skill when:
10
11	- Building a loop where an agent proposes edits to its own harness, prompts, context playbook, or workflow based on mined failure patterns
12	- Designing meta-level search over harness or scaffold code: meta-agent search, tree search over workflow graphs, evolutionary program search with an LLM mutation operator
13	- Choosing acceptance criteria for any self-modifying agent system
14	- Evolving the mechanism that manages context (a skill, playbook, or context function) rather than hand-editing the context artifact
15	- Diagnosing a degenerating self-improvement loop: reward hacking, diversity collapse, context collapse, or silent stagnation
16	- Deciding which level of the optimization ladder (prompt, context, workflow, harness code, optimizer code) a recurring failure should be fixed at
17
18	Do not activate this skill for adjacent work owned by other skills:
19
20	- Governance of a single autonomous loop that does not modify itself: locked and editable surfaces, durable logs, rollback, novelty gates, PR preparation, and human approval boundaries belong to `harness-engineering`. That skill defines the control surfaces; this skill defines what happens when the surfaces themselves become the optimization target.
21	- Building the evaluator, regression suite, or quality gates that score candidates: `evaluation`.
22	- LLM-as-judge design, pairwise comparison, and bias mitigation for candidate scoring: `advanced-evaluation`.
23	- One-shot token efficiency, masking, or caching without an improvement loop: `context-optimization`.
24	- Remote sandboxes and background execution infrastructure for running the loop: `hosted-agents`.
25	- Whether to build the loop at all, pipeline shape, and cost estimation: `project-development`.
26
27	## Core Concepts
28
29	### The Optimization Ladder
30
31	Published self-improvement systems target progressively deeper objects:
32
33	\| Rung \| Optimized object \| Example systems \|
34	\| --- \| --- \| --- \|
35	\| 1 \| Instruction prompts \| Promptbreeder, GEPA \|
36	\| 2 \| Structured context \| ACE (playbook of itemized bullets) \|
37	\| 3 \| Context mechanism \| MCE (the skill that produces context) \|
38	\| 4 \| Workflow graph \| ADAS, AFlow \|
39	\| 5 \| Harness code \| Self-Harness (bounded self-edits), Meta-Harness, Darwin Godel Machine \|
40	\| 6 \| Optimizer code \| STOP (the improver improves the improver) \|
41
42	Each rung up buys a larger design space and more leverage, and costs more per evaluation with a larger surface for gaming. Fix a recurring failure at the lowest rung that can express the fix. A stale-library failure is a context fix, not a workflow rewrite. Only move up when failure clusters at the current rung persist across candidates.
43
44	### The Loop Is Not the Intelligence
45
46	Recursive structure amplifies base-model quality in both directions. The same recursive improver design that compounds gains with a strong model degrades mean performance across iterations with weaker models (claim-self-improvement-capability-threshold). Two implications:
47
48	1. Run a capability validation before enabling recursion: a fixed number of iterations on a held-out task set, with the loop counted as net-negative if the trajectory declines.
49	2. Detect the degenerate stagnation mode. A broken improver that silently returns its input unchanged looks like stability in aggregate metrics. Track edit diffs, not only scores.
50
51	### The Outside-the-Loop Invariant
52
53	The evaluator, its instrumentation, permission control, and budget enforcem

People who liked this also viewed

6 picks

Type