The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied raw ideas and reads local profile files (~/.claude/x/voice.md) to generate drafted content. This ingestion surface is well-guarded by instructions that require the agent to present all results to the user for confirmation. The skill is explicitly forbidden from publishing content autonomously.
[OBFUSCATION]: The count.py script includes a regular expression that uses Unicode Tag characters (U+E0020–U+E007F). These are used for the legitimate technical purpose of correctly identifying and weighting multi-part emoji sequences (like regional indicator flags) to match X's weighted character counting standard. This is not a malicious obfuscation technique.
[COMMAND_EXECUTION]: The skill executes local Python scripts (count.py and hookscore.py) and references a sibling project script (weights.py) to perform linting and scoring of post drafts. These operations are conducted within the local agent environment on provided text inputs and do not involve remote code execution or unauthorized access.

x-post