2.5k
GitHub Stars
2
Bundled Files
2 months ago
Catalog Refreshed
3 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill openclaw/skills --skill knowledgebase-share- _meta.json292 B
- SKILL.md2.4 KB
Overview
This skill operates a multi-agent shared knowledge layer using a single GitHub repository as the canonical store. It standardizes where agents keep private drafts, how reusable content is promoted to shared knowledge, and enforces branch + PR workflows to keep shared material stable and auditable. Use it to set up, sync, and maintain consistent knowledge practices across agents.
How this skill works
The skill reads a configuration file (references/kb-config.json) to discover the canonical repo URL, local path, branch names, and folder roots before performing any action. It manages repository workflows: cloning, creating per-agent branches, syncing with rebase, committing private notes, and preparing PRs to promote content into the protected main branch. It also provides clear conflict-resolution and no-secrets rules so knowledge remains safe and consistent.
When to use it
- Setting up a shared knowledge repository for multiple agents or teams
- Creating per-agent private note folders and enforcing branch isolation
- Promoting reusable knowledge into a stable, reviewed main branch via PRs
- Synchronizing local agent branches with the latest shared content
- Resolving merge conflicts while preserving both versions for safe refactor
Best practices
- Always load configuration from references/kb-config.json or an explicit override; never hardcode repo URLs or secrets
- Pull and rebase before writing: git pull --rebase origin <branch> to minimize conflicts
- Keep private drafts under private/<agent>/ and only copy refined material into shared/ for PRs
- Promote to main only via PR and review; avoid force-pushing main
- When resolving conflicts, preserve both versions first, then refactor to a single clean artifact
Example use cases
- An agent creates daily research notes in private/agentA/, then promotes a distilled playbook into shared/ via a PR
- Onboarding a new agent: clone repo, create agent/<name> branch, and seed private folder templates
- A team synchronizes knowledge: fetch origin, rebase agent branch onto latest main, then merge reviewed changes
- Resolving a conflict where two agents edited the same playbook—both versions are preserved, then merged into a consolidated playbook in shared/
FAQ
Always read references/kb-config.json first. It contains repo_url, local_path, branch, private_root, and shared_root; use an explicit override only when provided.
Can agents push directly to main?
No. Shared knowledge enters main only via pull requests. Treat main as stable and avoid force-pushes or direct edits to preserve auditability.