self-improve

Name: self-improve
Author: jmstar85

151stars12forks

How to install self-improve?

jmstar85/oh-my-githubcopilot/self-improve

$npx -y skills add jmstar85/oh-my-githubcopilot --skill self-improve

Installs into the current project.

Summary

Autonomous evolutionary code improvement engine with tournament selection. Activate when user says: self-improve, self improve, evolve code, improve iteratively, tournament, benchmark loop, optimize code.

Files · 1

View on GitHub

SKILL.md

1	# Self-Improvement Orchestrator
2
3	Autonomous loop controller for evolutionary code improvement. Manages the full lifecycle: setup, research, planning, execution, tournament selection, history recording, and stop-condition evaluation.
4
5	## When to Use
6	- You want to iteratively improve a codebase toward a measurable benchmark goal
7	- Optimization tasks: performance, bundle size, test coverage, accuracy
8	- Code quality improvement with measurable metrics
9
10	## When NOT to Use
11	- No measurable benchmark available
12	- One-shot fix or feature request → use `/omg-autopilot`
13	- Manual, interactive coding → use `/ralph`
14
15	## Autonomous Execution Policy
16
17	NEVER stop or pause to ask the user during the improvement loop. Once the gate check passes and the loop begins, run fully autonomously until a stop condition is met.
18
19	- Do not ask for confirmation between iterations
20	- On agent failure: retry once, then skip and continue
21	- On all plans rejected: log it, continue to next iteration
22	- The only things that stop the loop are the stop conditions in Step 11
23
24	## State Tracking
25
26	All state lives under `.omg/self-improve/`:
27
28	```
29	.omg/self-improve/
30	├── config/
31	│ ├── settings.json # agents, benchmark, thresholds, sealed_files
32	│ ├── goal.md # Improvement objective + target metric
33	│ ├── harness.md # Guardrail rules (H001/H002/H003)
34	│ └── idea.md # User experiment ideas
35	├── state/
36	│ ├── agent-settings.json # iterations, best_score, status, counters
37	│ ├── iteration_state.json # Within-iteration progress (resumability)
38	│ ├── research_briefs/ # Research output per round
39	│ ├── iteration_history/ # Full history per round
40	│ ├── merge_reports/ # Tournament results
41	│ └── plan_archive/ # Archived plans (permanent)
42	├── plans/ # Active plans (current round)
43	└── tracking/
44	├── raw_data.json # All candidate scores
45	├── baseline.json # Initial benchmark score
46	└── events.json # Config changes
47	```
48
49	## Agent Mapping
50
51	\| Step \| Role \| Agent \| Purpose \|
52	\|------\|------\|-------\|---------\|
53	\| Research \| Codebase analysis \| @explore + @architect \| Hypothesis generation \|
54	\| Planning \| Hypothesis → plan \| @planner \| Structured plan per agent \|
55	\| Architecture Review \| 6-point review \| @architect \| Advisory review \|
56	\| Critic Review \| Harness enforcement \| @critic \| Approve/reject plans \|
57	\| Execution \| Implement + benchmark \| @executor \| Implement plan faithfully \|
58	\| Git Operations \| Merge/tag/PR \| @git-master \| Atomic merge operations \|
59
60	## Setup Phase
61
62	### Interactive Hook Protocol (Setup Only)
63
64	MANDATORY: Use `vscode_askQuestions` for all user decisions during setup (when available).
65	If `vscode_askQuestions` is NOT available (e.g., Copilot CLI), present numbered options in markdown and ask the user to respond with a number or freeform text.
66	Once the improvement loop begins (after gate passes), run fully autonomously — no hooks.
67
68	### When to Fire Hooks
69	\| Trigger Point \| Question Type \|
70	\|---------------\|---------------\|
71	\| Step 1: Repo path \| Confirm or select target repo \|
72	\| Step 4: Trust confirmation \| Confirm benchmark execution consent \|
73	\| Step 5: Goal interview \| Socratic goal-setting questions (Objective, Metric, Target, Scope) \|
74	\| Step 7: Harness rules \| Confirm/customize harness rules \|
75
76	1. Check if target repo path exists. If not configured:
77	HOOK: Ask user for target repo path via `vscode_askQuestions`:
78	```
79	header: "self-improve-target"
80	question: "Which repository should be improved?"
81	options: [
82	{ label: "[current workspace path]", recommended: true },
83	{ label: "Other — I'll specify the path" }
84	]
85	allowFreeformInput: true
86	```
87	2. Create `.omg/self-improve/` directory structure.
88	3. Read `agent-settings.json`. Check setup flags.
89	4. HOOK: Trust confirmation (mandatory) via `vscode_askQuestions`:
90	```
91	header: "self-improve-trust"
92	question: "Self-improve will run benchmarks and modify code in [repo path]. This includes running tests, builds, and git operations. Confirm?"
93	options: [
94	{ label: "Yes, I trust this — proceed", recommended: true },
95	{ label: "No, abort" }
96	]
97	allowFreeformInput: false
98	```
99	- If declined: abort immediately.
100	- Record consent: `trust_confirmed: true`
101	5. If goal not set → HOOK: Socratic goal interview via sequential `vscode_askQuestions`:
102	- Q1: Objective ("What do you want to improve?") with options: performance, test coverage, bundle size, code quality, accuracy, other
103	- Q2: Metric ("How should we measure it?") with context-derived options
104	- Q3: Target ("What's the target value?") with freeform
105	- Q4: Sco

People who liked this also viewed

6 picks

Type

self-improve

151stars12forks

View on GitHub

How to install self-improve?

jmstar85/oh-my-githubcopilot/self-improve

$npx -y skills add jmstar85/oh-my-githubcopilot --skill self-improve

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Self-Improvement Orchestrator
2
3	Autonomous loop controller for evolutionary code improvement. Manages the full lifecycle: setup, research, planning, execution, tournament selection, history recording, and stop-condition evaluation.
4
5	## When to Use
6	- You want to iteratively improve a codebase toward a measurable benchmark goal
7	- Optimization tasks: performance, bundle size, test coverage, accuracy
8	- Code quality improvement with measurable metrics
9
10	## When NOT to Use
11	- No measurable benchmark available
12	- One-shot fix or feature request → use `/omg-autopilot`
13	- Manual, interactive coding → use `/ralph`
14
15	## Autonomous Execution Policy
16
17	NEVER stop or pause to ask the user during the improvement loop. Once the gate check passes and the loop begins, run fully autonomously until a stop condition is met.
18
19	- Do not ask for confirmation between iterations
20	- On agent failure: retry once, then skip and continue
21	- On all plans rejected: log it, continue to next iteration
22	- The only things that stop the loop are the stop conditions in Step 11
23
24	## State Tracking
25
26	All state lives under `.omg/self-improve/`:
27
28	```
29	.omg/self-improve/
30	├── config/
31	│ ├── settings.json # agents, benchmark, thresholds, sealed_files
32	│ ├── goal.md # Improvement objective + target metric
33	│ ├── harness.md # Guardrail rules (H001/H002/H003)
34	│ └── idea.md # User experiment ideas
35	├── state/
36	│ ├── agent-settings.json # iterations, best_score, status, counters
37	│ ├── iteration_state.json # Within-iteration progress (resumability)
38	│ ├── research_briefs/ # Research output per round
39	│ ├── iteration_history/ # Full history per round
40	│ ├── merge_reports/ # Tournament results
41	│ └── plan_archive/ # Archived plans (permanent)
42	├── plans/ # Active plans (current round)
43	└── tracking/
44	├── raw_data.json # All candidate scores
45	├── baseline.json # Initial benchmark score
46	└── events.json # Config changes
47	```
48
49	## Agent Mapping
50
51	\| Step \| Role \| Agent \| Purpose \|
52	\|------\|------\|-------\|---------\|
53	\| Research \| Codebase analysis \| @explore + @architect \| Hypothesis generation \|
54	\| Planning \| Hypothesis → plan \| @planner \| Structured plan per agent \|
55	\| Architecture Review \| 6-point review \| @architect \| Advisory review \|
56	\| Critic Review \| Harness enforcement \| @critic \| Approve/reject plans \|
57	\| Execution \| Implement + benchmark \| @executor \| Implement plan faithfully \|
58	\| Git Operations \| Merge/tag/PR \| @git-master \| Atomic merge operations \|
59
60	## Setup Phase
61
62	### Interactive Hook Protocol (Setup Only)
63
64	MANDATORY: Use `vscode_askQuestions` for all user decisions during setup (when available).
65	If `vscode_askQuestions` is NOT available (e.g., Copilot CLI), present numbered options in markdown and ask the user to respond with a number or freeform text.
66	Once the improvement loop begins (after gate passes), run fully autonomously — no hooks.
67
68	### When to Fire Hooks
69	\| Trigger Point \| Question Type \|
70	\|---------------\|---------------\|
71	\| Step 1: Repo path \| Confirm or select target repo \|
72	\| Step 4: Trust confirmation \| Confirm benchmark execution consent \|
73	\| Step 5: Goal interview \| Socratic goal-setting questions (Objective, Metric, Target, Scope) \|
74	\| Step 7: Harness rules \| Confirm/customize harness rules \|
75
76	1. Check if target repo path exists. If not configured:
77	HOOK: Ask user for target repo path via `vscode_askQuestions`:
78	```
79	header: "self-improve-target"
80	question: "Which repository should be improved?"
81	options: [
82	{ label: "[current workspace path]", recommended: true },
83	{ label: "Other — I'll specify the path" }
84	]
85	allowFreeformInput: true
86	```
87	2. Create `.omg/self-improve/` directory structure.
88	3. Read `agent-settings.json`. Check setup flags.
89	4. HOOK: Trust confirmation (mandatory) via `vscode_askQuestions`:
90	```
91	header: "self-improve-trust"
92	question: "Self-improve will run benchmarks and modify code in [repo path]. This includes running tests, builds, and git operations. Confirm?"
93	options: [
94	{ label: "Yes, I trust this — proceed", recommended: true },
95	{ label: "No, abort" }
96	]
97	allowFreeformInput: false
98	```
99	- If declined: abort immediately.
100	- Record consent: `trust_confirmed: true`
101	5. If goal not set → HOOK: Socratic goal interview via sequential `vscode_askQuestions`:
102	- Q1: Objective ("What do you want to improve?") with options: performance, test coverage, bundle size, code quality, accuracy, other
103	- Q2: Metric ("How should we measure it?") with context-derived options
104	- Q3: Target ("What's the target value?") with freeform
105	- Q4: Sco

People who liked this also viewed

6 picks

Type