Security audit
Security checks for vulnerabilities and agentic risk
The skill appears locally scoped and non-destructive, but its executable workflow does not actually perform the promised translation and polishing.
Review this skill before installing if you need real translation and publishing output: the included workflow appears to create a simulated result rather than a polished zh-CN article. Also remember that drafts are saved under ~/.openclaw by default, so choose a different output directory or clean up files when working with sensitive content.
The manifest says the skill will polish a technical blog into an 800–1000 word zh-CN article while preserving terms and code. In code, the polish_and_translate step simply wraps the original content in a fixed string and writes it out, without performing any restructuring, translation, or length enforcement.
The markdown instructions say outputs should use yymmddhhmm-* naming and include a subject-derived filename. The code sets ts=$(date +"%H%M") and writes only ${ts}-polished.md, which directly contradicts the documented naming behavior.
The description promises consistent hero and per-section image prompts for the user's blog, implying prompts should reflect the draft topic and section count. The code always emits exactly three fixed filenames and three generic prompts, regardless of the blog subject, requested style/background/aspect ratio, or number of sections.
The skill persists user-provided or derived content to a stable workspace path under the user's home directory, which can leave sensitive draft material on disk beyond the current session. In a content-processing skill, this increases privacy and data-retention risk, especially if drafts contain unpublished, confidential, or proprietary information.
- code blocks and commands (no translation inside code fences)
- inline code spans (backticks) as-is unless obviously wrong
- **Images**: generate a **hero prompt** + **one prompt per section**, all in the same style/tone
- **Save**: write everything to `~/.openclaw/workspace/contentPolished/` using `yymmddhhmm-*` naming
## When to Use This Skill
The instruction says that when the user asks to polish a technical blog, translate it to Simplified Chinese, and create images, the agent 'MUST use this skill,' and the skill is explicitly fixed to zh-CN. This enforces a specific locale/language behavior rather than offering a choice, which matches the language-policy violation category.
The documented workflow explicitly creates timestamped files in a persistent workspace, which can accumulate historical copies of user content and metadata over time. This increases exposure if the workstation is shared, backed up to external systems, or later compromised.
draftPath = input.draftPath || "~/.openclaw/workspace/contentDraft/latestDraft.md"outputDir = input.outputDir || "~/.openclaw/workspace/contentPolished/"ts in yymmddhhmm format (Linux/macOS):date +"%y%m%d%H"
Output size or generation rate is not bounded. Unbounded output enables denial-of-service through resource exhaustion, log flooding, or context-window stuffing.
- `skills`, `plugin`, `OpenClaw`, `ClawHub`, `CLI`, `Markdown`, `YAML`
- If a technical term has a standard Chinese translation but the English term is common in your audience, keep the English term and (optionally) add a short Chinese clarification once on first use
### Step 5 — Enforce Length (1000–1200 Words) Without Cutting Meaning
Target final length 800–1000 words (counting Chinese words approximately by rough equivalence).
The save step formalizes durable storage of the polished article and related image filenames, creating a predictable artifact trail on disk. Predictable, persistent output paths make it easier for other local processes or users to discover unpublished content, which is a real confidentiality concern even though it is not remote code execution.
1. Decide `subject`:
- `subject = input.subject || slugify(title)` (lowercase, hyphens)
2. Write the polished markdown:
- `{{outputDir}}/{{ts}}-{{subject}}.md`
3. Determine image filenames:
- Hero: `{{ts}}-main.png`
The documentation describes a careful content-processing workflow with preservation of code fences, inline code, and technical terms through translation. The implementation merely loads the entire file with cat and passes raw content onward, with no evidence of the promised exact-preservation or structured parsing behavior.
No suspicious patterns detected.