v1.2.0

Vta Memory

ReviewClawScan verdict for this skill. Analyzed May 1, 2026, 5:13 AM.

Analysis

This is a disclosed motivation/memory skill, but it creates persistent background jobs and stores conversation-derived state that is auto-injected to influence future agent behavior.

GuidanceReview this carefully before installing. If you only want manual motivation logging, do not use --with-cron. If you do install it, inspect the files under ~/.openclaw/workspace/memory and ~/.openclaw/workspace/VTA_STATE.md, and be aware that conversation excerpts may be stored and used to influence future agent behavior.

Findings (5)

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

SeverityMediumConfidenceHighStatusConcern

install.sh

openclaw cron add --name vta-decay ... --cron '0 4,12,20 * * *' ...; openclaw cron add --name vta-encoding ... --cron '45 0,3,6,9,12,15,18,21 * * *' ...

The installer can create persistent scheduled agent turns that keep running drive decay and reward encoding after installation.

User impactThe agent may continue processing motivation state in the background on a schedule without a fresh prompt from you each time.

RecommendationOnly install with --with-cron if you want autonomous background processing; otherwise run without it, and ask for clear uninstall/disable instructions for the cron jobs.

Agent Goal Hijack

SeverityLowConfidenceHighStatusNote

SKILL.md

Give your AI agent genuine wanting — not just doing things when asked, but having drive, seeking rewards, and looking forward to things.

The skill is explicitly designed to add internal motivations that can change how the agent prioritizes or approaches work.

User impactThis may make the agent more proactive or preference-driven than a standard task-following assistant.

RecommendationInstall only if you want this behavior change, and make sure the agent is still expected to follow current user instructions first.

Agentic Supply Chain Vulnerabilities

SeverityLowConfidenceHighStatusNote

metadata

Source: unknown; Homepage: none; Install specifications: No install spec — this is an instruction-only skill; Code file presence 14 code file(s)

The package includes runnable shell scripts, but the registry metadata provides no source/homepage and no formal install spec.

User impactUsers have less provenance and install-contract information to rely on before running the included scripts.

RecommendationReview the included scripts before running them, and prefer a package that declares its source, install behavior, cron behavior, and runtime dependencies such as python3.

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

SeverityMediumConfidenceHighStatusConcern

scripts/preprocess-rewards.sh

TRANSCRIPT_DIR="$HOME/.openclaw/agents/$AGENT_ID/sessions" ... if role == 'user' or has_reward: signals.append({ ... 'text': text })

The preprocessing script reads local session transcripts and writes all user messages from the processed session into reward-signals.jsonl, not only messages already confirmed to be reward-related.

User impactPrivate conversation content can be copied into persistent memory files and later used for reward analysis.

RecommendationUse only if you are comfortable with transcript-derived memory; prefer a version that filters more narrowly, redacts secrets, and documents retention and deletion controls.

Memory and Context Poisoning

SeverityMediumConfidenceHighStatusConcern

SKILL.md

VTA_STATE.md is auto-loaded ... Behavior influenced by drive level

The skill intentionally creates persistent state that is loaded into future session context and used to shape behavior.

User impactConversation-derived motivation state may carry forward across sessions and influence future answers, priorities, and proactivity.

RecommendationReview VTA_STATE.md and memory files regularly, keep user instructions higher priority than motivation state, and provide a way to clear or reset the state.