v1.0.0

KJ's Self-Improving Proactive Agent

BenignClawScan verdict for this skill. Analyzed May 1, 2026, 8:36 AM.

Analysis

This is a coherent instruction-only skill for persistent agent memory and proactive follow-up, with clear privacy boundaries but some local state and provenance details users should notice.

GuidanceInstall this only if you want the agent to keep local long-term memory and proactive task state. Review the `~/self-improving/` and `~/proactivity/` files occasionally, avoid putting secrets in them, and confirm the publisher/source if the metadata mismatch matters to you.

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.

Abnormal behavior control

Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.

Rogue Agents

SeverityLowConfidenceHighStatusNote

heartbeat-rules.md

Good heartbeat checks - promised follow-ups that are due - stale blockers that may now be unblocked - deadlines or reviews approaching soon

The skill includes proactive heartbeat behavior that can surface follow-ups from stored state. It is disclosed and bounded, not hidden background execution.

User impactThe agent may proactively remind, draft, or recommend next steps based on saved state.

RecommendationUse it only if you want proactive follow-up, and keep the approval boundaries for external actions in place.

Agentic Supply Chain Vulnerabilities

SeverityInfoConfidenceMediumStatusNote

_meta.json

"ownerId": "kn7c88h93b3n1bsepkes2zfggx82z7f4", "slug": "self-improving-proactive-agent"

The packaged metadata does not match the supplied registry metadata exactly, which lists a different owner ID and registry slug. This is a provenance/coherence note, not evidence of unsafe behavior.

User impactThe package identity is slightly inconsistent, so users may want to confirm they are installing the intended skill.

RecommendationVerify the publisher and homepage before relying on the skill, especially because it affects persistent agent behavior.

Sensitive data protection

Checks for exposed credentials, poisoned memory or context, unclear communication boundaries, or sensitive data that could leave the user's control.

Memory and Context Poisoning

SeverityLowConfidenceHighStatusNote

state.md

Durable learning → `~/self-improving/` Use for: confirmed preferences, reusable lessons, domain habits, project-specific patterns

The skill intentionally stores durable user and project preferences that can be reused in later tasks, so incorrect or stale entries could steer future behavior.

User impactSaved memory can make the agent more useful, but it can also preserve mistakes or private project details until reviewed or removed.

RecommendationPeriodically review and prune the local memory files, and do not store secrets or sensitive credentials in them.