Skillv1.2.0

ClawScan security

Context Hygiene · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 28, 2026, 5:19 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: An instruction-only guide for pruning agent workspace context that is internally consistent with its stated purpose, but it recommends deleting/modifying workspace files so review and backups are advised before use.
Guidance: This is a coherent, instruction-only policy for keeping agent context small; it does not request credentials or install software. However, it prescribes deleting and editing workspace files (pruning MEMORY.md, removing stale files, editing AGENTS.md). Before enabling an agent to follow these rules: (1) Back up your workspace (or test in a disposable/sandbox workspace) so deletions are reversible; (2) Review any automated deletion rules and ensure they target only the intended files; (3) If the agent will be given write/delete permissions, restrict those to a limited directory or require explicit user confirmation before destructive actions; (4) Manually set USER.md timezone as recommended rather than relying on the agent to infer it; (5) If you want to adopt the policy, verify the canonical source (the SKILL cites a GitHub repo) before copying files into your workspace. Overall this skill is coherent and low-risk conceptually, but file-modifying behaviors are operationally impactful and should be controlled.

Review Dimensions

Purpose & Capability: okThe name/description (context hygiene) matches the SKILL.md instructions. The skill is instruction-only and asks for no binaries, env vars, or installs — everything requested is proportionate to a policy document about trimming agent context.
Instruction Scope: noteThe instructions explicitly tell agents how to read, prune, collapse, and delete workspace files (MEMORY.md, daily memory files, HEARTBEAT.md, AGENTS.md etc.), recommend spawning sub-agents, and give shell examples (| head, jq). This behavior is coherent with a 'hygiene' policy, but it grants the agent authority to modify or remove user files. That is expected for the purpose but operationally impactful — users should ensure backups and limit write permissions if they don't want automatic deletions.
Install Mechanism: okNo install spec and no code files (instruction-only). This is the lowest-risk install profile: nothing is downloaded or written by the skill itself.
Credentials: okThe skill requires no environment variables or credentials. It mentions external systems (e.g., JIRA profiles and system locale) only as context for why a USER.md timezone is useful; it does not request access to those services or credentials.
Persistence & Privilege: noteThe skill does not set always:true and asks for no special privileges. However, it recommends edits to AGENTS.md and copying a CONTEXT.md into the workspace — actions that, if performed, will make the hygiene rules persistent. The skill itself does not automatically persist or modify other skills.