Obsidian Skills
kepano/obsidian-skillsSix skills by Obsidian's Steph Ango (kepano) for Obsidian Markdown, Bases, JSON Canvas, the Obsidian CLI, Defuddle and Knap.
Forty skills that let an agent build and maintain a linked Obsidian vault from your notes, code and past agent sessions, then answer questions with citations.
A large, well-documented pack where all state is plain Markdown on your disk; expect to spend time on setup and vault hygiene.
obsidian-wiki follows the compile-once knowledge base idea: the agent reads your sources and writes interlinked Markdown pages into an Obsidian vault you own, instead of re-answering from scratch each time. The core commands are wiki-setup, wiki-capture (save what you just worked out), wiki-ingest (documents, PDFs, transcripts, URLs), wiki-query (answers with wikilink citations), wiki-update (distill the repo you are in), plus wiki-lint, wiki-dedup, wiki-digest and wiki-narrate for upkeep and summaries.
A family of history-ingest skills mines past sessions from Claude Code, Codex, Copilot, Hermes, OpenClaw and Pi, and session-search helps find a conversation you cannot name. There are 40 skills averaging about 14 KB, some with a reference file, and an optional Python CLI installed with pip that also registers two session hooks (skip with --no-hooks). The Claude Code plugin install ships skills only. The README says no API key or network is needed because the vault is Markdown on disk. The pack also bundles a copy of a skill-creator toolkit with eval scripts. Quality depends on how well your agent follows long procedural prompts.
40 skills averaging 14 KB with some reference files, a Python CLI and a bundled eval toolkit, but few scripts tied to the wiki skills themselves.
Counted from the repository file tree. How we score depth.
A set of Obsidian knowledge-vault skills that ingest your documents and past agent session logs (Claude, Codex, Copilot and others) into a markdown wiki. The pip-installed CLI registers session hooks by default unless you pass --no-hooks, the setup skill can fetch and register a Stop hook from the main branch, and an optional daily maintenance job installs a launchd, systemd or cron schedule. History ingestion reads other tools' local transcripts, and vault git sync pushes to a remote you choose.
obsidian-wiki setup (pip/uv/pipx path) also registers two documented session hooks (session start and stop capture) by default; the README says to pass --no-hooks to skip them, and the wiki-setup skill asks before installing the Stop hook. The hook and CLI code are not in our files.README.md:48pip install obsidian-wiki (or uv/pipx), which provides the CLI and hooks; the plugin route ships skills only.README.md:41obsidian-wiki sync stages, commits and pushes the vault to a git remote you configure (opt-in)..skills/wiki-setup/SKILL.md:360The repository (556 files) on Oct 29, 2026. Not a full review: 470 of 556 files were not downloaded (large repository; we reviewed the skill folders); we read 0 of 10 code files; it relies on code outside the repository: the obsidian-wiki PyPI package / CLI, scripts/daily-update.sh, hooks/wiki-stop-capture.sh and the session hooks (470 of 556 files were not downloaded); qmd CLI; history-ingest skills read local session stores of other agents (~/.claude/projects, Claude Desktop sessions, ~/.codex, Copilot, Hermes, OpenClaw, Pi) and distill them into the vault; they tell the agent to redact secrets, and this reading was not exercised.
NVIDIA SkillSpector static scan: 51 pattern hits (Rogue Agent 12, Agent Snooping 11, Supply Chain 10, Privilege Escalation 5, Dangerous Code Execution 3, Excessive Agency 3). The 51 hits include 10 'Concealed Executable Artifact' matches on the bundled skill-creator scripts, an 'ignore previous instructions' phrase inside wiki-ingest's trust-boundary section that tells the agent to treat sources as data, agent-config path mentions in history-ingest docs, and the skill-creator's os.environ copy for claude -p. Hook install, scheduler and curl steps are recorded as flags.
A review, not a certification. Skills change after we check them — read the files before installing.
MIT — commercial use allowed.
Not really. Ingest and upkeep skills run on request and ask before including sensitive material; there is no autonomous loop.
Six skills by Obsidian's Steph Ango (kepano) for Obsidian Markdown, Bases, JSON Canvas, the Obsidian CLI, Defuddle and Knap.
Cross-agent persistent memory for Claude Code, Codex, OpenClaw, OpenCode and DeepSeek Harness: Markdown files as the source of truth, with a Milvus search index.
MCP server plus hooks that keep bulky tool output out of the context window, index it in local SQLite, and restore session state after compaction.
It is a set of 40 agent skills that build and maintain an Obsidian vault as a linked knowledge base. The agent ingests documents and chat history, then answers questions with wikilink citations.
Run pip install obsidian-wiki then obsidian-wiki setup --vault ~/brain, or install it as a Claude Code plugin with /plugin marketplace add Ar9av/obsidian-wiki. Then tell your agent to set up the wiki.
The README says there is no network, no API key and no vector store; the vault is local Markdown. Your agent's own model provider still sees whatever the agent reads.