stark-ai-de/agent-skills / plugins/stark-ai-developer/skills/karakeep-knowledge
karakeep-knowledge
UnverifiedSet up or use a local persistent Karakeep knowledge collection through its official MCP server. Use when configuring Karakeep for coding agents, checking an existing connection, checking project readiness, saving factual session reports, synchronizing approved canonical documents and separate drafts, maintaining approved source-backed aggregates, curating lists and tags, or consulting saved web sources for development work.
references/hermes.md
View on GitHub# Hermes installation and session checks
Use this adapter when connecting Hermes to the same knowledge workflow. It is one
portable skill, not a separate Hermes copy of the reconciliation logic. Installing
or loading it grants no collection access or repository-import permission.
## Install the complete package
Resolve the actual execution host and active profile's `HERMES_HOME` first. Hermes
normally uses `~/.hermes`; named profiles have their own configuration and skill
directory. Preserve existing configuration, profile isolation and local edits.
Install the entire reviewed `karakeep-knowledge/` directory, including `scripts/`,
`scripts/knowledge/`, `references/`, `assets/` and `agents/`. A lone `SKILL.md` or
`knowledge.mjs` cannot run the helper: its imported modules must be present. Use an
immutable source revision and record the installed revision/digest privately.
Use one installation, chosen by the existing configuration owner:
- Profile-local: the complete folder at `$HERMES_HOME/skills/karakeep-knowledge`.
- Shared read-only package: the complete folder under an existing external skills
root, with that root added to the profile's `skills.external_dirs`.
Do not blindly overwrite an existing installation or enable every unrelated skill
from a different host. Check for a profile-local or trusted project skill with the
same name: a trusted project copy can override it; conflicting profile-local and
external copies can instead make native loading refuse an ambiguous name. Resolve
the intended source before claiming that the shared revision is in use. Preserve
both copies; explicitly select one installation root or an unambiguous categorized
path in that profile's configuration. A container image can install this
nonsecret payload at build time; it must not bake in credentials or private policy.
## Use Hermes' native auto-load configuration
Merge the appropriate values from [the example](../assets/hermes-config.example.yaml)
into the active profile's `config.yaml`, preserving unrelated lists and settings.
For a profile-local installation, omit `external_dirs`. The example path is a
placeholder, not an existing installation. Do not replace the whole YAML file.
`skills.auto_load` loads this skill's instructions at the start of a normal Hermes
agent session. The agent then invokes the Node `check` at its first substantive Git
repository work. This is agent behavior, not a deterministic command hook, daemon,
cron job or post-merge pipeline. Hermes' own disabled-skill settings, ignored rules
and sessions without the skills toolset/context can suppress auto-loading; internal
delegation/review agents need not inherit it. Verify the actual target session.
Do not treat `~/.codex/AGENTS.md` or a file at `$HERMES_HOME/AGENTS.md` as a global
Hermes startup mechanism. Hermes project context has its own discovery/precedence.
No edit to `SOUL.md` or the agent's identity is needed. For an older Hermes without
native auto-load, invoke the skill once manually and adopt the existing
[project instruction](../assets/project-instructions.md) in the context file that
the installed Hermes actually loads. Do not upgrade the runtime silently.
## Execute on the correct host
The helper requires Node.js 24.18+ and Git on Linux/POSIX. Resolve the installed
skill directory shown by Hermes and run:
```sh
node /installed/karakeep-knowledge/scripts/knowledge.mjs check --repo /absolute/repository
```
Hermes' terminal backend must be able to read that same package and repository.
Host-side discovery does not prove availability inside a remote/Docker terminal
backend. Use the selected profile's private descriptor, credential references and
shared collection state under the [helper contract](helper.md). The existing helper
and MCP launcher accept loopback origins only: another container's or host's
`localhost` does not reach the owner's service. Report missing connectivity; do not
weaken the origin check, expose the service, copy host credentials into an image or
invent cross-host writer coordination. The designated writer and shared state must
remain consistent with the adopted policy.
## Qualification and disablement
1. In the actual profile, verify that Hermes discovers the intended installed skill
and can read its references and every imported Node module.
2. Start a fresh ordinary session in a Git repository. Observe the native auto-load
and the agent's real read-only `check` invocation. A loaded prompt alone proves
instruction discovery, not execution or a working Karakeep connection.
3. Verify `sync: not_enabled` without mutation for an unadopted repository and continued
repository work when the configured service is unavailable. Qualify authorized
reads/writes separately using the existing setup/sync acceptance cases.
Removing `auto_load` or disabling the skill stops automatic instruction loading; it
does not revoke retained API/MCP credentials. To enforce a write stop, the connection
owner must revoke/detach the attachment-owned write capability. Preserve knowledge,
other attachments and unrelated credentials. Keep runtime, CI, published-image and
live-session evidence distinct.
## Checked upstream contract
Checked 2026-10-16 against Hermes revision
`f97608f178d1ffeca59860195ab7da295f7c8e5f`:
[native defaults](https://github.com/NousResearch/hermes-agent/blob/f97608f178d1ffeca59860195ab7da295f7c8e5f/hermes_cli/config_defaults.py),
[session auto-load](https://github.com/NousResearch/hermes-agent/blob/f97608f178d1ffeca59860195ab7da295f7c8e5f/agent/system_prompt.py)
and [skill discovery](https://github.com/NousResearch/hermes-agent/blob/f97608f178d1ffeca59860195ab7da295f7c8e5f/agent/skill_utils.py).
The [skills guide](https://hermes-agent.nousresearch.com/docs/user-guide/features/skills/)
and [context guide](https://hermes-agent.nousresearch.com/docs/user-guide/features/context-files/)
describe the current user surface; recheck when changing the pinned runtime.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |