Self-Improving + Proactive Agent

PassAudited by ClawScan on May 10, 2026.

Overview

This instruction-only skill is coherent and disclosed, but it intentionally creates persistent local memory and workspace steering that can affect future agent behavior.

Install this only if you want the agent to keep local cross-session learning about your preferences and corrections. Review the AGENTS.md, SOUL.md, HEARTBEAT.md, and ~/self-improving/ changes, keep sensitive data out of memory, and approve the optional Proactivity install only after reviewing that skill separately.

Findings (3)

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.

NoteHigh Confidence

ASI06: Memory and Context Poisoning

What this means

The agent may remember and apply your preferences, corrections, and project patterns in later tasks.

Why it was flagged

The skill intentionally loads persistent learned memory into future sessions, so incorrect, overly broad, or sensitive entries could influence later agent behavior.

Skill content

### On Session Start
1. Load memory.md (HOT tier)
2. Check index.md for context hints
3. If project detected  preload relevant namespace

Recommendation

Use the audit and forget commands, avoid storing secrets or sensitive personal data, and periodically review ~/self-improving/ for entries you do not want retained.

NoteHigh Confidence

ASI10: Rogue Agents

What this means

The skill can continue influencing future sessions through workspace heartbeat and memory files.

Why it was flagged

The setup can add recurring heartbeat steering that keeps the memory system maintained after initial setup.

Skill content

Add this section to your `HEARTBEAT.md` ... Use `~/self-improving/heartbeat-state.md` for last-run markers and action notes

Recommendation

Only add the HEARTBEAT/AGENTS/SOUL steering if you want persistent self-improvement behavior, and remove those snippets if you later disable the skill.

NoteHigh Confidence

ASI04: Agentic Supply Chain Vulnerabilities

What this means

Approving the companion install adds another skill with its own instructions and risks.

Why it was flagged

The skill may install a separate Proactivity skill that is outside the artifacts reviewed here, although it requires explicit user approval.

Skill content

Only install it after the user explicitly agrees.

If the user agrees:

1. Run `clawhub install proactivity`

Recommendation

Review the Proactivity skill separately before agreeing to install or activate it.