self-improving-agent

Name: self-improving-agent
Author: Tyuts

46stars5forks

How to install self-improving-agent?

tyuts/xiaobai-skills/self-improving-agent

$npx -y skills add Tyuts/xiaobai-skills --skill self-improving-agent

Installs into the current project.

Summary

Repo-local meta-skill for improving Xiaobai Skills over time. Use after significant project changes, user feedback, install failures, promotion experiments, or repeated workflow friction to update AGENTS.md, capture project-specific gotchas, or propose new repo-local skills. Do n

Files · 1

View on GitHub

SKILL.md

1	# Self-Improving Agent
2
3	This is a repo-local skill for Xiaobai Skills. It helps the project learn from
4	real work without turning every task into expensive reflection.
5
6	Use it only when one of these happens:
7
8	- A user reports an install or curation failure.
9	- A promotion experiment teaches something reusable.
10	- The recommended starter stack changes.
11	- The agent repeatedly searches for the same repo knowledge.
12	- A workflow repeats enough to deserve a repo-local skill.
13	- A release reveals missing docs, checks, or restore steps.
14
15	Do not use it after every small edit.
16
17	## Step 1: Diagnose The Learning
18
19	Ask:
20
21	1. Did the repo structure or command flow change?
22	2. Did the user correct a behavior that should persist?
23	3. Did an install command fail in a way future agents should know?
24	4. Did a starter-stack decision prove good or bad?
25	5. Did a repeated workflow emerge?
26
27	If the answer is no, make no changes.
28
29	## Step 2: Choose The Right Artifact
30
31	- Update `AGENTS.md` for short repo rules, commands, gotchas, or decision
32	history.
33	- Update `README.md` for user-facing install or positioning changes.
34	- Update `ROADMAP.md` for future work.
35	- Update `docs/promotion-kit.md` for launch and marketing learnings.
36	- Create `skills/<name>/SKILL.md` only for repeatable multi-step workflows.
37
38	Prefer one concise update over several files.
39
40	## Step 3: Update AGENTS.md Safely
41
42	When editing `AGENTS.md`:
43
44	- Keep entries short and repo-specific.
45	- Add dated decision notes only when they explain future behavior.
46	- Remove or revise stale notes instead of appending contradictions.
47	- Keep the file readable for every future session.
48	- Never add secrets, payment details, tokens, or private user data.
49
50	## Step 4: Create Repo-Local Skills Sparingly
51
52	Create a new skill only when the workflow is:
53
54	- repeated
55	- multi-step
56	- specific enough to benefit from written procedure
57	- useful for future maintainers
58
59	Use:
60
61	```text
62	skills/<skill-name>/SKILL.md
63	```
64
65	The skill must include:
66
67	- clear `name`
68	- clear `description`
69	- trigger conditions
70	- steps
71	- output contract
72
73	## Step 5: Validate
74
75	Before finishing:
76
77	1. Re-read changed files.
78	2. Check for contradictions with existing docs.
79	3. Keep project-level instructions conservative.
80	4. If code or scripts changed, run the relevant validation.
81	5. Summarize the learning in the final response.
82
83	## Anti-Patterns
84
85	- Turning one-off preferences into permanent rules.
86	- Creating global skills for Xiaobai-specific workflows.
87	- Recording private sponsorship/payment data in repo docs.
88	- Enabling broad self-reflection on every task.
89	- Creating verbose AGENTS.md notes that cost more than they save.

People who liked this also viewed

6 picks

Type

self-improving-agent

46stars5forks

View on GitHub

How to install self-improving-agent?

tyuts/xiaobai-skills/self-improving-agent

$npx -y skills add Tyuts/xiaobai-skills --skill self-improving-agent

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Self-Improving Agent
2
3	This is a repo-local skill for Xiaobai Skills. It helps the project learn from
4	real work without turning every task into expensive reflection.
5
6	Use it only when one of these happens:
7
8	- A user reports an install or curation failure.
9	- A promotion experiment teaches something reusable.
10	- The recommended starter stack changes.
11	- The agent repeatedly searches for the same repo knowledge.
12	- A workflow repeats enough to deserve a repo-local skill.
13	- A release reveals missing docs, checks, or restore steps.
14
15	Do not use it after every small edit.
16
17	## Step 1: Diagnose The Learning
18
19	Ask:
20
21	1. Did the repo structure or command flow change?
22	2. Did the user correct a behavior that should persist?
23	3. Did an install command fail in a way future agents should know?
24	4. Did a starter-stack decision prove good or bad?
25	5. Did a repeated workflow emerge?
26
27	If the answer is no, make no changes.
28
29	## Step 2: Choose The Right Artifact
30
31	- Update `AGENTS.md` for short repo rules, commands, gotchas, or decision
32	history.
33	- Update `README.md` for user-facing install or positioning changes.
34	- Update `ROADMAP.md` for future work.
35	- Update `docs/promotion-kit.md` for launch and marketing learnings.
36	- Create `skills/<name>/SKILL.md` only for repeatable multi-step workflows.
37
38	Prefer one concise update over several files.
39
40	## Step 3: Update AGENTS.md Safely
41
42	When editing `AGENTS.md`:
43
44	- Keep entries short and repo-specific.
45	- Add dated decision notes only when they explain future behavior.
46	- Remove or revise stale notes instead of appending contradictions.
47	- Keep the file readable for every future session.
48	- Never add secrets, payment details, tokens, or private user data.
49
50	## Step 4: Create Repo-Local Skills Sparingly
51
52	Create a new skill only when the workflow is:
53
54	- repeated
55	- multi-step
56	- specific enough to benefit from written procedure
57	- useful for future maintainers
58
59	Use:
60
61	```text
62	skills/<skill-name>/SKILL.md
63	```
64
65	The skill must include:
66
67	- clear `name`
68	- clear `description`
69	- trigger conditions
70	- steps
71	- output contract
72
73	## Step 5: Validate
74
75	Before finishing:
76
77	1. Re-read changed files.
78	2. Check for contradictions with existing docs.
79	3. Keep project-level instructions conservative.
80	4. If code or scripts changed, run the relevant validation.
81	5. Summarize the learning in the final response.
82
83	## Anti-Patterns
84
85	- Turning one-off preferences into permanent rules.
86	- Creating global skills for Xiaobai-specific workflows.
87	- Recording private sponsorship/payment data in repo docs.
88	- Enabling broad self-reflection on every task.
89	- Creating verbose AGENTS.md notes that cost more than they save.

People who liked this also viewed

6 picks

Type