Stego Text — Steganographic Message Encoding for AI Agents

Encode and decode hidden messages within natural-looking text using steganographic techniques. Use when asked to hide a message in text, encode a secret mess...

This skill is coherent for its stated purpose: it teaches multiple ways to hide and recover messages in ordinary text and requires no extra permissions. Before installing, consider these practical cautions: (1) the encoding methods can be repurposed to hide sensitive information in otherwise-normal outputs, so avoid enabling the skill for agents that handle secrets or private data; (2) if you allow autonomous invocation, monitor outputs for hidden channels (acrostics, unusual word choices, sentence-length patterns) because the skill makes it easy to embed exfiltration channels; (3) test the skill in a safe sandbox with non-sensitive data to verify decode/verification steps work as expected; and (4) if you need to limit risk, disallow autonomous invocation for this skill or require human approval before it runs. If you'd like, I can suggest guardrails or detection checks you can apply to agent outputs to detect common steganographic patterns described here.

Type: OpenClaw Skill Name: stego-text Version: 1.0.1 The skill bundle provides instructions and reference material for performing various forms of textual steganography, such as acrostics, word-count mapping (Structural Harmonic), and statistical anomalies (Resonance Encoding). It contains no executable code, no network or file system access, and no instructions that would lead to data exfiltration or unauthorized system access. The content is entirely focused on the stated purpose of encoding and decoding hidden messages within natural-looking text, primarily through files like SKILL.md and encoding-methods.md.