Self Improving Agent Local

PassAudited by ClawScan on May 10, 2026.

Overview

This appears to be a disclosed local learning/reminder skill, but its persistent memory files, optional hooks, and cross-session features should be reviewed before use.

This skill is reasonable for local self-improvement workflows. Before installing, verify the package/source identity, enable hooks only if you want recurring reminders, and review any learned content before promoting it into persistent agent instruction files. Avoid logging secrets or sensitive transcripts.

Findings (4)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

Future sessions may follow saved learnings or instructions, including incorrect or overly broad guidance.

Why it was flagged

Promoted learnings become persistent context that future agents may rely on, so mistaken or sensitive entries could influence later behavior.

Skill content
Broadly applicable learning | Promote to `CLAUDE.md`, `AGENTS.md`, and/or `.github/copilot-instructions.md` ... OpenClaw injects these files into every session
Recommendation

Review and approve promotions to CLAUDE.md, AGENTS.md, SOUL.md, TOOLS.md, or Copilot instructions; redact secrets and avoid storing untrusted instructions as authoritative memory.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

Every enabled session may receive extra self-improvement instructions, which can slightly steer the agent and consume context.

Why it was flagged

When enabled, the hook persistently injects a self-improvement reminder into agent bootstrap context.

Skill content
Fires on `agent:bootstrap` ... Adds a reminder block ... Prompts the agent to log corrections, errors, and discoveries
Recommendation

Enable the hook only if you want recurring reminders, inspect the hook content first, and disable it if it becomes intrusive.

What this means

Session transcripts or learnings could be shared across agent sessions, and background sub-agents could act on delegated tasks.

Why it was flagged

The documentation describes cross-session transcript access, messaging, and sub-agent spawning for sharing learnings.

Skill content
sessions_history — Read another session's transcript ... sessions_send — Send a learning to another session ... sessions_spawn — Spawn a sub-agent for background work
Recommendation

Use cross-session tools only within trusted workspaces, avoid reading or forwarding sensitive transcripts without approval, and confirm before spawning sub-agents.

What this means

A user following the instructions could install a different source or package than the artifact being reviewed.

Why it was flagged

The install guidance references an external repository and a package name that differ from the evaluated registry slug `self-improving-agent-local`.

Skill content
clawdhub install self-improving-agent ... git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
Recommendation

Verify the package slug, owner, and repository before installing, and prefer the reviewed registry artifact when possible.