id	ai-knowledge-base.skill.knowledge-authoring
kind	skill
name	ai-knowledge-base-knowledge-authoring
description	Author or update tool-neutral knowledge packs in ai-knowledge-base.
tags	["knowledge","authoring","repository"]

Knowledge Authoring

Use this skill when creating or updating content under knowledge/ in the ai-knowledge-base repository.

Authoring Flow

Inspect the target pack first.
- Read manifest.yaml.
- Read nearby entries of the same kind.
- Preserve existing naming, frontmatter, and tone unless there is a clear reason to change them.
Choose the right entry kind.
- Use skill for reusable procedures, workflows, checklists, viewpoints, or output formats.
- Use agent for a specialist role that can operate in an isolated context.
- Use rule for always-on instructions.
- Use prompt for reusable prompt or command-style entry points.
Add or update the manifest entry.
- Entry id format is <pack>.<kind>.<name>.
- Skill paths are directories: skills/<name>.
- Agent, rule, and prompt paths are files: agents/<name>.md, rules/<name>.md, or prompts/<name>.md.
Add or update the entry source.
- Skills place their body in skills/<name>/SKILL.md.
- Other entry kinds are single markdown files.
- YAML frontmatter must include id, kind, name, and description.
- The frontmatter id must exactly match the manifest entry id.
Keep the content tool-neutral.
- Describe concepts and workflows independently of Claude, Codex, Gemini, or another target unless the entry is intentionally target-specific.
- Use tools.<target> only for target-specific frontmatter or enablement.
Validate before finishing.
- Run go test ./... from knit/.
- If local sandboxing blocks the default Go caches, run with GOCACHE and GOMODCACHE pointed at writable directories inside the repository or /tmp.

Quality Bar

A skill should say when to use it, what context to inspect, what steps to follow, and what output or verification is expected.
An agent should define its role, scope, inputs, review criteria, and output format.
A rule should be short enough to stay always-on without distracting from the active task.
A prompt should be directly executable by a user or assistant and should make required inputs clear.
Prefer concrete repository paths and id examples over generic prose.
Avoid duplicating large sections from another entry. Extract shared guidance into a separate skill only when it will be reused.

Más de este repositorio

mismo repositorio

go-test-authoring

theoden9014/ai-knowledge-base

A workflow for authoring Go tests. Starting from the table-driven test scaffold generated by gotests, it runs the cycle of test design -> adding test cases -> running -> fixing defects. Use when: (1) invoked via /go-test-authoring, (2) asked to "write tests" or "add tests" for Go, (3) asked to implement tests for a Go function or method.

2026-06-262

git-commit

theoden9014/ai-knowledge-base

Git commit with Conventional Commits format. Use when user asks to commit changes, says "commit", "/commit", or after completing code changes when explicitly asked to commit.

2026-06-222

pr-create

theoden9014/ai-knowledge-base

Create GitHub Pull Request with project's PR template. Use when user asks to create PR, says "create PR", "/pr-create", or after completing implementation when asked to open a pull request.

2026-06-222

pr-review-fix

theoden9014/ai-knowledge-base

Address PR review feedback. Collect both unresolved inline comments and review-body comments on a GitHub PR, build a fix plan and turn it into tasks, then work through them one by one (assess -> fix -> commit -> push -> reply). Use when invoked as /pr-review-fix, when asked to "address the review comments" or "respond to PR feedback", or for follow-up work after a PR review.

2026-06-222

structure-behavior-design-conceptual-modeling

theoden9014/ai-knowledge-base

Internal Structure-Behavior Design skill for conceptual modeling and structure design. Use only as part of the Structure-Behavior Design workflow.

2026-05-222

structure-behavior-design-interface-design

theoden9014/ai-knowledge-base

Internal Structure-Behavior Design skill for interface, signature, and contract design. Use only as part of the Structure-Behavior Design workflow.

2026-05-222

Knowledge Authoring

Use this skill when creating or updating content under knowledge/ in the ai-knowledge-base repository.

Authoring Flow

Inspect the target pack first.

Read manifest.yaml.
Read nearby entries of the same kind.
Preserve existing naming, frontmatter, and tone unless there is a clear reason to change them.

Choose the right entry kind.

Use skill for reusable procedures, workflows, checklists, viewpoints, or output formats.
Use agent for a specialist role that can operate in an isolated context.
Use rule for always-on instructions.
Use prompt for reusable prompt or command-style entry points.

Add or update the manifest entry.

Entry id format is <pack>.<kind>.<name>.
Skill paths are directories: skills/<name>.
Agent, rule, and prompt paths are files: agents/<name>.md, rules/<name>.md, or prompts/<name>.md.

Add or update the entry source.

Skills place their body in skills/<name>/SKILL.md.
Other entry kinds are single markdown files.
YAML frontmatter must include id, kind, name, and description.
The frontmatter id must exactly match the manifest entry id.

Keep the content tool-neutral.

Describe concepts and workflows independently of Claude, Codex, Gemini, or another target unless the entry is intentionally target-specific.
Use tools.<target> only for target-specific frontmatter or enablement.

Validate before finishing.

Run go test ./... from knit/.
If local sandboxing blocks the default Go caches, run with GOCACHE and GOMODCACHE pointed at writable directories inside the repository or /tmp.

Quality Bar

A skill should say when to use it, what context to inspect, what steps to follow, and what output or verification is expected.

An agent should define its role, scope, inputs, review criteria, and output format.

A rule should be short enough to stay always-on without distracting from the active task.

A prompt should be directly executable by a user or assistant and should make required inputs clear.

Prefer concrete repository paths and id examples over generic prose.

Avoid duplicating large sections from another entry. Extract shared guidance into a separate skill only when it will be reused.