Context Slimming

PassAudited by VirusTotal on May 10, 2026.

Overview

Type: OpenClaw Skill Name: context-slimming Version: 1.0.2 The 'context-slimming' skill is a utility designed to optimize OpenClaw workspace markdown files (e.g., AGENTS.md, SOUL.md) to reduce token usage. It provides a structured workflow for auditing file sizes, removing redundant templates, and refactoring detailed rules into subdirectories. The skill uses standard shell commands (wc, cat) and git for version control. It includes specific guidance on maintaining security constraints while condensing text. No evidence of malicious intent, data exfiltration, or unauthorized execution was found; the behavior is entirely consistent with its stated purpose of workspace optimization.

Findings (0)

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.

ConcernHigh Confidence
ASI08: Cascading Failures
What this means

The agent's future behavior could change across sessions if important instructions or safeguards are removed or over-compressed.

Why it was flagged

The skill is designed to edit the files that are injected into every future agent session. A bad compression decision could remove safety rules, user preferences, or operational context and affect later tasks.

Skill content
OpenClaw 将所有顶层 `*.md` 文件自动注入到每轮对话的系统提示中。 ... **顶层注入文件原则**:每文件 ≤ 1,500 bytes。
Recommendation

Require an explicit user-approved plan, show a diff before applying changes, keep backups, and validate behavior after edits before replacing the original context files.

What this means

Private profile details, memory, or credential-location notes could be mishandled, removed, or retained in injected context in ways the user did not intend.

Why it was flagged

The workflow includes editing persistent user profile, memory, and tool/credential-location context, but does not define privacy boundaries, retention rules, or approval requirements for what gets kept, removed, or rewritten.

Skill content
USER.md # 只留:用户画像摘要 + 沟通策略 ... MEMORY.md # 只留:关键决策摘要 + 错误模式索引 ... TOOLS.md # 只留:凭证位置 + 关键配置
Recommendation

Treat USER.md, MEMORY.md, and TOOLS.md as sensitive; avoid storing secrets, preserve safety-critical instructions, and ask the user before changing or deleting persistent memory or credential references.

What this means

Unrelated local changes could be committed together with the context-slimming edits.

Why it was flagged

The default submission step stages all workspace changes with git add -A and commits them, without first requiring a diff review or limiting the commit to files modified by this skill.

Skill content
cd /path/to/workspace
git add -A
git commit -m "workspace: context slimming — structure + dedup + optional classical"
Recommendation

Add an explicit `git diff` review step, ask for user confirmation before committing, and stage only the intended files.

What this means

Credential locations or operational configuration may remain in always-injected context or be rewritten incorrectly.

Why it was flagged

The skill may inspect or rewrite documentation about credential locations and key configuration. This is purpose-aligned for slimming TOOLS.md, but it involves sensitive operational context.

Skill content
TOOLS.md # 只留:凭证位置 + 关键配置
Recommendation

Keep actual secrets out of injected markdown files and confirm that any credential references are minimal, necessary, and not copied into other files.