Context Pruner

Intelligent context window management by summarizing and removing redundant history. Helps agents maintain high performance in long-running threads.

Security Scan

The skill's declared purpose (pruning and summarizing conversation history) matches its minimal runtime instructions and it requests no credentials or installs, but the instructions are high-level and the package has no source/homepage so some operational details are missing.

Purpose & Capability

Name and description match the instructions: pruning, summarizing, and chunking conversational history. There are no unexpected required binaries, env vars, or config paths that would be incoherent with a context-management skill.

Instruction Scope

SKILL.md contains only high-level pruning steps (noise detection, fact distillation, chunking) and does not instruct reading unrelated files, network exfiltration, or credential access. However the instructions are vague about what gets deleted versus summarized and what heuristics are used, giving the skill broad discretion if implemented by an agent.

Install Mechanism

This is an instruction-only skill with no install spec or code files, so nothing will be written to disk by the skill itself. The README mentions a 'clawhub install' command but no install spec is present—this is a minor documentation incongruence rather than an evidence of risky installation behavior.

Credentials

The skill requests no environment variables, credentials, or config paths. There is no disproportionate access requested for unrelated services.

Persistence & Privilege

always is false and the skill is user-invocable. It does not request persistent system-wide changes or access to other skills' configs. Autonomous model invocation is allowed by default on the platform but that is not a red flag on its own here.

Assessment

This skill appears internally consistent with its stated goal and requests no credentials, but it is minimal and undocumented (no source or homepage). Before installing: 1) Ask the provider how it decides what to delete vs summarize and whether actions are reversible (retain backups). 2) Test on non-sensitive conversations first. 3) Verify what 'clawhub install' would do in your environment. 4) If you need stronger assurances, request source code or documentation from the publisher so you can confirm it doesn't transmit or permanently erase data you care about.

Save tokens and keep focus sharp. This skill manages the "bloat" of long sessions.

Pruning Protocol

1. Noise Detection: Filter "Acknowledge" messages and filler words.
2. Fact Distillation: Extract raw info and discard the conversational fluff.
3. Chunking: Break long transcripts into searchable summaries.

Installation

clawhub install context-pruner