Text

Transform, format, and process text with patterns for writing, data cleaning, localization, citations, and copywriting.

This skill is a coherent, instruction-only text-processing guide and appears to do what it says. Before using: (1) Review any shell/python commands and run them on copies of your data (avoid sed -i on originals). (2) Be cautious when processing .env, config, or log files — they often contain secrets; do not run extraction commands on those files unless you intend to and understand the risk. (3) Some examples rely on tools (python3, jq, yq, csvkit); ensure you have the intended versions and prefer non-destructive flags or dry-run options. (4) If you plan to allow an automated agent to run these instructions, restrict it to safe test data or a sandbox to avoid accidental data modification or exposure.

Type: OpenClaw Skill Name: text Version: 1.0.0 The skill bundle is primarily instructional and provides numerous benign text processing examples. However, the `data.md` file includes shell commands that execute arbitrary Python code (`python3 -c "..."`) and commands to parse `.env` files (`grep ... .env`). While the examples provided are benign and align with the stated purpose of text processing, these capabilities introduce a vulnerability risk. The `python3 -c` command could be leveraged for Remote Code Execution (RCE) if an AI agent were to construct or receive malicious input for it. The `.env` file parsing, while a legitimate data extraction task, could expose sensitive environment variables if applied to a privileged file by a compromised agent or malicious prompt. These are risky capabilities that could allow attacks, classifying the bundle as suspicious rather than benign, but without clear evidence of intentional malicious design.