skill-improvement

Improve an existing skill in diegocanepa/agent-skills based on conversation learnings and submit the changes via a Pull Request.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "skill-improvement" with this command: npx skills add diegocanepa/agent-skills/diegocanepa-agent-skills-skill-improvement

🔧 Skill Improvement

Identify gaps or enhancement opportunities in an existing skill from the current conversation and contribute an improved version to the shared repository.

When to Use

Use this skill when:

  • A skill didn't behave as expected and you've identified the root cause.
  • The conversation reveals a missing step, edge case, or unclear instruction in a skill.
  • A skill should reference another skill or MCP that wasn't mentioned.
  • The user says "improve this skill", "update this skill", or "this skill needs X".

Workflow

Step 1: Identify the Target Skill

Clarify with the user (if not already clear):

  • Which skill needs improvement? (e.g. swe-orchestrator, vcs-commit)
  • What specifically should be improved? (missing step, wrong reference, unclear instruction, new edge case)

Step 2: Read the Current Skill

Read the current SKILL.md of the target skill from skills/<skill-name>/SKILL.md.

Step 3: Extract Improvements from the Conversation

Review the conversation and identify:

  • What failed or was unclear in the skill's current instructions.
  • What new patterns, steps, or guardrails should be added.
  • Whether any references to other skills, MCPs, or tools are missing.
  • Whether the description frontmatter still accurately describes when to use the skill.

Present a diff summary to the user before making changes:

Proposed Improvements:
- [ ] [Section/Line]: [What changes and why]
- [ ] [Section/Line]: [What changes and why]

Step 4: Apply the Improvements

Edit the skills/<skill-name>/SKILL.md file with the identified improvements.

Quality check after editing:

Improvement Review:
- [ ] The `description` frontmatter still accurately triggers the skill
- [ ] New instructions are self-contained and clear
- [ ] No existing correct behavior was accidentally removed
- [ ] References to other skills use relative paths (e.g. [vcs-commit](../vcs-commit/SKILL.md))
- [ ] Change is focused — does not introduce unrelated refactoring

If any check fails → revise → re-check.

Step 5: Create the Branch

Use vcs-branch to create a dedicated branch:

fix/improve-skill-<skill-name>

or for larger additions:

feat/enhance-skill-<skill-name>

Step 6: Commit the Changes

Use vcs-commit to commit:

git add skills/<skill-name>/SKILL.md
git commit -m "fix(skills): improve <skill-name> - <short description of what changed>"
git push -u origin fix/improve-skill-<skill-name>

Step 7: Create the Pull Request

Use vcs-change-request and the GitHub MCP to open a PR:

  • Owner: diegocanepa
  • Repo: agent-skills
  • URL: https://github.com/diegocanepa/agent-skills
  • Head branch: fix/improve-skill-<skill-name>
  • Base branch: main

PR description template:

## Summary

Improves the **<skill-name>** skill.

## Problem

<What was unclear, missing, or broken in the current version.>

## Changes

- [Section changed]: <What was added/modified and why>
- [Section changed]: <What was added/modified and why>

## Conversation Context

This improvement was identified from a conversation about: <topic>.

## Checklist
- [ ] Existing correct behavior preserved
- [ ] Instructions are clearer and more actionable
- [ ] No unrelated changes introduced

Step 8: MITM — Confirm with User

"I've applied the improvements to <skill-name> and the PR is ready at https://github.com/diegocanepa/agent-skills. Does the diff look correct?"

Feedback Loop: If the USER wants to refine further → adjust → amend/push → re-present.


Notes

  • Keep improvements focused. One PR per skill, one concern per PR.
  • Never rewrite a skill entirely unless the whole structure is broken. Prefer targeted edits.
  • If the improvement is substantial and changes the skill's purpose, consider using add-skill to create a new variant instead.
  • The description frontmatter is critical — if the improvement changes when the skill applies, update it.

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Automation

mermaid-diagrams

No summary provided by upstream source.

Repository SourceNeeds Review
Automation

add-skill

No summary provided by upstream source.

Repository SourceNeeds Review
Automation

vcs-branch

No summary provided by upstream source.

Repository SourceNeeds Review