- Home
- Skills
- Agents & Orchestration
- Offline Reflection & Knowledge-Graph Memory by Saki-AI
Offline Reflection & Knowledge-Graph Memory by Saki-AI
Review daily workspace changes and turn evidence into durable notes, typed entities and knowledge-graph relations. Includes a local Node.js scanner. Free from the creator of Saki-AI and Saki-Panel.
Free
Works with the AI Agents you already use
Offline Reflection & Knowledge-Graph Memory by Saki-AI
Example session with this skill installed
In a synthetic workspace, SESSION-STATE.md says: Goal: Ship the sample project. Decision: Keep the public API stable. Next step: Review the release checklist. The only other file is src/status.js containing module.exports = { status: "ready-for-review" }. Both were modified today. Scan the workspace, then propose a durable memory and one knowledge-graph relation. Do not write files.
- Read your context and instructions
- Compiled the offline reflection
Local scan
The bundled scanner was run against the two-file synthetic fixture. It returned changedFileCount: 2 and digests for:
SESSION-STATE.mdsrc/status.js
Agent-produced consolidation proposal
{
"memoryCandidates": [
{
"text": "Keep the sample project public API stable during release work.",
"source": "SESSION-STATE.md",
"confidence": 0.95
}
],
"entities": [
{"id": "sample-project", "type": "project", "name": "Sample project"},
{"id": "release-checklist", "type": "goal", "name": "Review the release checklist"}
],
"relations": [
{
"from": "sample-project",
"type": "next_step",
"to": "release-checklist",
"source": "SESSION-STATE.md"
}
]
}
The digest comes from the scanner; the proposed memory and graph are an illustrative agent interpretation. Confidence is a review aid, not a statistical guarantee. The ready-for-review value is temporary session state and should not become an enduring personal preference. No files were written.
Next step: review the candidates, merge duplicates with existing memory, then request a memory update if desired.
Connects securely to your tools. The creator never sees your data.
About this skill
Carry useful project knowledge into the next session
A free workflow extracted from Saki-AI for daily reviews, session wrap-ups and persistent project memory. Replace an unstructured pile of notes with evidence-based memory candidates and typed relationships.
Included
- A SKILL.md protocol for reviewing changes and distilling
memoryCandidates,entitiesandrelations. scripts/offline-reflection-scanner.cjs, a zero-dependency Node.js scanner that collects bounded excerpts from supported text files modified on the local calendar day.- Guidance for updating MEMORY.md and SESSION-STATE.md after reviewing the candidates.
Clear division of work
The scanner reads files and emits a JSON digest. Your agent interprets that digest and any user-selected session notes, proposes memories and writes approved updates. The script alone does not extract a knowledge graph or rewrite your memory files.
Run node scripts/offline-reflection-scanner.cjs <workspace> against a workspace you choose. Common dependency, build and hidden directories are excluded. The digest is bounded and is not a complete audit of every file.
Free under Apache-2.0. No npm packages or network calls in the scanner.
Explore both projects
This free skill is shared by the creator of Saki-AI and Saki-Panel. Visit both repositories, try the projects, and consider giving both a GitHub star to support continued development.
- Saki-AI: https://github.com/EthanChan050430/Saki-AI
- Saki-Panel: https://github.com/EthanChan050430/Saki-Panel
Uses the local calendar day and filesystem modification times. Includes at most 10 file excerpts from a bounded candidate scan; ignores common dependency/build/hidden directories and large files. It is not a complete audit. Memory extraction and optional writes require an agent.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 15 days ago
- Free with an Agensi account
Requires
Permissions
Reads supported text files in a user-selected workspace. The scanner only emits JSON; memory writes are performed by the agent when requested. It makes no network calls. Sending digests to an API-based agent depends on your chosen agent.